CN110490305A - Machine learning model processing method and node based on block chain network - Google Patents
Machine learning model processing method and node based on block chain network Download PDFInfo
- Publication number
- CN110490305A CN110490305A CN201910776690.2A CN201910776690A CN110490305A CN 110490305 A CN110490305 A CN 110490305A CN 201910776690 A CN201910776690 A CN 201910776690A CN 110490305 A CN110490305 A CN 110490305A
- Authority
- CN
- China
- Prior art keywords
- machine learning
- learning model
- block chain
- node
- chain network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/23—Updating
- G06F16/2365—Ensuring data consistency and integrity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computer Security & Cryptography (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of machine learning model processing method, node and storage mediums based on block chain network;Machine learning model processing method includes: the machine learning model for obtaining party in request's node and sending, and the machine learning model is stored to block chain network;In response to the inquiry request to the machine learning model that participant node is sent, the machine learning model is sent to the participant node, so that the participant node is updated the machine learning model according to the training data of itself;The updated machine learning model that the participant node is sent is obtained, is known together to the updated machine learning model;When knowing together successfully, the updated machine learning model is stored to the block chain network.By means of the invention it is possible in the machine learning model Training scene participated in many ways, the safety of training for promotion data, while the treatment effeciency of hoisting machine learning model.
Description
Technical field
The present invention relates to artificial intelligence and block chain technology more particularly to a kind of machine learning moulds based on block chain network
The node and storage medium of type processing method, block chain network.
Background technique
Artificial intelligence (AI, Artificial Intelligence) is to utilize digital computer or digital computer control
Machine simulation, extension and the intelligence for extending people of system, perception environment obtain knowledge and the reason using Knowledge Acquirement optimum
By, method, technology and application system.Machine learning (ML, Machine Learning) is a multi-field cross discipline, is people
The core of work intelligence, is the fundamental way for making computer have intelligence, and application spreads every field.
Can training data be the Key Asset that determines machine learning model and effectively use, for different company and enterprises
Or for department, government organs, it is generally owned by the unique training data of oneself, according to the difference in source, these training datas are past
Toward being distributed in independent " data silo ".It is usually to establish Data Mart or big in the scheme that the relevant technologies provide
Data platform, shared on it and training data of trading, so as to a large amount of training datas according to separate sources to machine learning
Model is updated.
But the application scenarios increasingly complex of machine learning model, such as might have multiple business divisions and participate in instruction
Practice process, the training program that the relevant technologies provide is unable to satisfy the demand of safety and efficiency.
Summary of the invention
The embodiment of the present invention provides a kind of machine learning model processing method based on block chain network, node and storage and is situated between
Matter can guarantee safety and efficiency in the machine learning model Training scene participated in many ways.
The technical solution of the embodiment of the present invention is achieved in that
The embodiment of the present invention provides a kind of machine learning model processing method based on block chain network, comprising:
The machine learning model that party in request's node is sent is obtained, the machine learning model is stored to block chain network;
In response to the inquiry request to the machine learning model that participant node is sent, by the machine learning model
It is sent to the participant node, so that training data of the participant node according to itself, to the machine learning model
It is updated;
The updated machine learning model that the participant node is sent is obtained, to the updated engineering
Model is practised to know together;
When knowing together successfully, the updated machine learning model is stored to the block chain network.
It is in the above scheme, described to store the machine learning model to block chain network, comprising:
The block chain network includes for the unified block chain for storing different machine learning models, by the engineering
Model storage is practised into the block chain of the block chain network;Alternatively,
The block chain network includes multiple channels corresponding with different types of machine learning model, by the engineering
Model storage is practised into the block chain in corresponding types channel.
The embodiment of the present invention provides a kind of node of block chain network, comprising:
Model obtains module, for obtaining the machine learning model of party in request's node transmission, by the machine learning model
It stores to block chain network;
Model transaction modules, the inquiry request to the machine learning model for being sent in response to participant node,
The machine learning model is sent to the participant node, so that training data of the participant node according to itself,
The machine learning model is updated;
Common recognition module, the updated machine learning model sent for obtaining the participant node, to update
The machine learning model afterwards is known together;
Module is updated storage, for when knowing together successfully, the updated machine learning model to be stored to the area
Block chain network.
In the above scheme, the model obtains module, is also used to:
The block chain network includes for the unified block chain for storing different machine learning models, by the engineering
Model storage is practised into the block chain of the block chain network;Alternatively,
The block chain network includes multiple channels corresponding with different types of machine learning model, by the engineering
Model storage is practised into the block chain in corresponding types channel.
The embodiment of the present invention provides a kind of node of block chain network, comprising:
Memory, for storing executable instruction;
Processor when for executing the executable instruction stored in the memory, is realized provided in an embodiment of the present invention
Machine learning model processing method.
The embodiment of the present invention provides a kind of storage medium, is stored with executable instruction, real when for causing processor to execute
Existing machine learning model processing method provided in an embodiment of the present invention.
The embodiment of the present invention has the advantages that
The embodiment of the present invention is using block chain network as the shared platform of machine learning model, from party in request's node to block
Chain network submits machine learning model, machine learning model is inquired from participant node to block chain network, in local to machine
Learning model is updated, and submits updated machine learning model to block chain network, can be in the machine participated in many ways
In learning model Training scene, the safety of training for promotion data, while the treatment effeciency of hoisting machine learning model.
Detailed description of the invention
Fig. 1 is that one of the machine learning model processing system provided in an embodiment of the present invention based on block chain network is optional
Configuration diagram;
Fig. 2 is an optional function structure schematic diagram of block chain network provided in an embodiment of the present invention;
Fig. 3 is an optional structural schematic diagram of the node of block chain network provided in an embodiment of the present invention;
Fig. 4 is that one of the machine learning model processing method provided in an embodiment of the present invention based on block chain network is optional
Flow diagram;
Fig. 5 is that another of the machine learning model processing method provided in an embodiment of the present invention based on block chain network can
The flow diagram of choosing;
Fig. 6 is that another of the machine learning model processing method provided in an embodiment of the present invention based on block chain network can
The flow diagram of choosing;
Fig. 7 is that another of the machine learning model processing method provided in an embodiment of the present invention based on block chain network can
The flow diagram of choosing.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with attached drawing to the present invention make into
It is described in detail to one step, described embodiment is not construed as limitation of the present invention, and those of ordinary skill in the art are not having
All other embodiment obtained under the premise of creative work is made, shall fall within the protection scope of the present invention.
In the following description, it is related to " some embodiments ", which depict the subsets of all possible embodiments, but can
To understand, " some embodiments " can be the same subsets or different subsets of all possible embodiments, and can not conflict
In the case where be combined with each other.
In the following description, related term " first second third " be only be the similar object of difference, no
Represent the particular sorted for being directed to object, it is possible to understand that ground, " first second third " can be interchanged specific in the case where permission
Sequence or precedence so that the embodiment of the present invention described herein can be other than illustrating herein or describing
Sequence is implemented.
Unless otherwise defined, all technical and scientific terms used herein and belong to technical field of the invention
The normally understood meaning of technical staff is identical.Term used herein is intended merely to the purpose of the description embodiment of the present invention,
It is not intended to limit the present invention.
Before the embodiment of the present invention is further elaborated, to noun involved in the embodiment of the present invention and term
It is illustrated, noun involved in the embodiment of the present invention and term are suitable for following explanation.
1) it trades (Transaction), is equal to computerese " affairs ", transaction includes needing to be submitted to block chain
The operation that network executes, not singly refers to the transaction in commercial environment, " hands in view of having used in block chain technology sanctified by usagely
Easily " this term, the embodiment of the present invention have followed this habit.
For example, deployment (Deploy) transaction is for the specified intelligent contract of node installation into block chain network and prepares
It is good called;Call (Invoke) transaction for the record by calling the additional transaction in block chain of intelligent contract, and to area
The slip condition database of block chain is operated, including updating operation (including the key assignments in increase, deletion and modification slip condition database
It is right) and inquiry operation (key-value pair i.e. in inquiry slip condition database).
2) block chain (Blockchain) is the storage knot of the encryption formed by block (Block), chain type transaction
Structure.
For example, the head of each block both may include the cryptographic Hash of All Activity in block, while also comprising previous
The cryptographic Hash of All Activity in block, to realize the anti-tamper and anti-counterfeiting traded in block based on cryptographic Hash;It is newly generated
Transaction is filled into block and after the common recognition of block chain network interior joint, can be appended to the tail portion of block chain to be formed
The growth of chain type.
3) block chain network (Blockchain Network), is included in the one of block chain for new block by way of common recognition
The set of the node of series.
4) account book (Ledger) is the system of block chain (also referred to as account book data) and the slip condition database synchronous with block chain
Claim.
Wherein, block chain is to be traded in the form of the file in file system to record;Slip condition database is with inhomogeneity
The form of key (Key) value (Value) pair of type records the transaction in block chain, for supporting quick to what is traded in block chain
Inquiry.
5) intelligent contract (Smart Contracts), also referred to as chain code (Chaincode) or application code, are deployed in area
Program in the node of block chain network, node execute the intelligent contract called in received transaction, carry out reconciliation database
The operation that key-value pair data is updated or inquires.
6) it knows together (Consensus), is a process in block chain network, for right between the multiple nodes being related to
Transaction in block is reached an agreement, and the block reached an agreement will be appended to the tail portion of block chain, and the mechanism for realizing common recognition includes
Proof of work (PoW, Proof of Work), equity prove (PoS, Proof of Stake), share authorisation verification (DPoS,
Delegated Proof-of-Stake), elapsed time amount prove (PoET, Proof of Elapsed Time) etc..
7) machine learning model refers to the rule of thumb existing structure of knowledge of data organization, to constantly improve self performance
Model, such as linear regression model (LRM), Random Forest model and neural network model etc..The training of machine learning model, needs
Training data is handled, and under certain constraint condition, learns the mathematical notation for meeting constraint condition optimal out.This
The machine learning model that text is related to mainly include model structure (such as neural network structure) and model parameter (such as mind
Through the connection weight parameter between each layer of network, the weight parameter of each characteristic item of logistic regression).
The exemplary application for illustrating block chain network provided in an embodiment of the present invention below is the present invention referring to Fig. 1, Fig. 1
The configuration diagram for the machine learning model processing system based on block chain network that embodiment provides, including block chain network
200 (illustrate and (illustrate including node 210-1 to node 210-3), authentication center 300, business division 400
Belong to the party in request's node 410 and its graphical interfaces 420 of business division 400) and business division 500 (illustrate ownership
In the participant node 510 and its graphical interfaces 520 of business division 500), wherein party in request's node 410 and participant node
510 be the client node in block chain network 200, and the content of the embodiment of the present invention is merely for convenience and purposes of illustration of shown in Fig. 1,
And it does not imply that party in request's node 410 and participant node 510 and exists independently of block chain network 200.In addition, authentication center can
Independently of block chain network, the certification node that also can be used as in block chain network exists.It is illustrated separately below.
The type of block chain network 200 is flexile, such as can be appointing in publicly-owned chain, privately owned chain or alliance's chain
It anticipates one kind.By taking publicly-owned chain as an example, the electronic equipment of any business division such as user terminal and server can not needed
Block chain network 200 is accessed in the case where authorization;By taking alliance's chain as an example, business division its electronics having under its command after being authorized is set
Standby (such as terminal/server) can access block chain network 200, and become the client node in block chain network 200.
It may be noted that ground, client node, which can be provided only, supports business division to initiate transaction (for example, storing for cochain
Data in data or inquiry chain) function, and for the function of the routine of block chain network 200 (primary) node 210, such as hereafter
Described ranking function, common recognition service and account book function etc., client node can be default or selective (for example, depending on
In the specific business need of business division) realize.It is thus possible to by the data of business division and business processing logic maximum journey
Degree moves in block chain network 200, realizes the credible of data and business procession by block chain network 200 and can chase after
It traces back.
Block chain network 200 is received from different business main body (such as business division 400 and business master shown in Fig. 1
Body 500) client node (for example, belonged to shown in Fig. 1 business division 400 party in request's node 410 and ownership
In the participant node 510 of business division 500) transaction submitted, transaction is executed to update account book or inquiry account book, and in visitor
The user interface of family end node is (for example, the graphical interfaces of the graphical interfaces 420 of party in request's node 410, participant node 510
520) display executes the various intermediate results or final result of transaction.
Block is hereafter illustrated by taking the processing for realizing machine learning model as an example by multiple business divisions access block chain network
The exemplary application of chain network.
Referring to Fig. 1, business division 400 accesses block chain network 200 by party in request's node 410, and business division 500 passes through
Participant node 510 accesses block chain network 200.Graphical interfaces of the business personnel of business division 400 in party in request's node 410
Machine learning model is inputted in 420, party in request's node 410 generates the corresponding transaction for submitting operation, reality is specified in transaction
The intelligent contract for now operation being submitted to need to call and the parameter transmitted to intelligent contract, transaction also carry party in request's node
The digital signature of 410 signatures is (for example, the private key in the digital certificate issued using authentication center 300, carries out the abstract of transaction
Encryption obtains), and transaction is broadcast to block chain network 200.
When receiving transaction in the node 210 in block chain network 200, the digital signature carried to transaction is verified,
After digital signature authentication success, according to the identity of the party in request's node 410 carried in transaction, whether party in request's node 410 is confirmed
It is that the verifying judgement of any one of digital signature and Authority Verification, which fails, all will lead to Fail Transaction with trading privilege.It tests
The digital signature of node 210 oneself is signed after demonstrate,proving successfully (for example, using the private key in the digital certificate of node 210-1 to transaction
Abstract encrypted to obtain), and continue broadcasted in block chain network 200.
After node 210 in block chain network 200 with ranking function receives the transaction being proved to be successful, transaction is filled
Into new block, and it is broadcast in block chain network 200 and the node of common recognition service is provided.
The node 210 of offer common recognition service in block chain network 200 carries out common recognition process to new block to reach an agreement,
New block is appended to the tail portion of block chain by the node 210 for providing account book function, and executes the transaction in new block: i.e. for mentioning
The corresponding key-value pair of machine learning model is added in the transaction for handing over machine learning model in account book database.
Similarly, the business personnel of business division 500 can input machine in the graphical interfaces 520 of participant node 510
Learning model/machine learning model inquiry request, participant node 510 are inquired according to machine learning model/machine learning model
Request generates corresponding update operation/inquiry operation transaction, and transaction is broadcast to block chain network 200.
After the node 210 in block chain network 200 is broadcasted, sorted and known together to the transaction, in block chain 200
The corresponding new block of the transaction is appended to the tail portion of block chain by the node 210 for providing account book function, and is executed in new block
Transaction: the corresponding key-value pair of machine learning model is inquired in the transaction for inquiring machine learning model from account book database, and
Return to query result;For submitting the transaction of updated machine learning model, machine learning model in account book database is updated
Corresponding key-value pair.It is worth noting that the updated machine learning model that participant node 510 is submitted, can be participation
Fang Jiedian utilizes the training data of itself, is updated to the machine learning model of inquiry.
It is to be appreciated that the data that business division can be inquired in block chain network 200/be updated, can pass through constraint
The permission for the transaction that business division can be initiated is realized, for example, initiating to submit machine learning model when business division 400 has
Transaction permission when, the business personnel of business division 400 can input machine in the graphical interfaces 420 of party in request's node 410
Device learning model, and generated by party in request's node 410 for submitting the transaction of machine learning model to be broadcast to block chain network 200
In, to add machine learning model in the account book of block chain network 200.
When business division 500 has the permission for initiating the transaction of update/inquiry machine learning model, business division 500
Business personnel update/inquiry request of machine learning model can be inputted in the graphical interfaces 520 of participant node 510,
And generated by participant node 510 and be broadcast in block chain network 200 for updating/inquiring the transaction of machine learning model, with
Machine learning model is updated in the account book of block chain network 200, or corresponding machine learning is obtained from block chain network 200
Model.
Illustrate the illustrative function structure of block chain network provided in an embodiment of the present invention below, referring to fig. 2, Fig. 2 is
The function structure schematic diagram of block chain network 200 provided in an embodiment of the present invention, including application layer 201, common recognition layer 202, network
Layer 203, data Layer 204 and resource layer 205, are illustrated separately below.
Resource layer 205 encapsulate the computing resource of each node 210 realized in block chain network 200, storage resource and
The communication resource, such as computer, the computing resource in server/cluster and cloud, storage resource and the communication resource be abstracted simultaneously
Unified interface is provided to data Layer 204 to shield the otherness for the bottom hardware for realizing resource layer 205.
Computing resource includes various forms of processors, such as central processing unit (CPU), application specific integrated circuit
(ASIC, Application Specific Integrated Circuit), specific integrated circuit and field programmable gate array
The various forms of processors of (FPGA, Field-Programmable Gate Array).
Storage resource includes various types of storage mediums such as various volatile memory and nonvolatile memory.Its
In, nonvolatile memory can be read-only memory (ROM, Read Only Memory), programmable read only memory
(PROM, Programmable Read-Only Memory).Volatile memory can be random access memory (RAM,
Random Access Memory), it is used as External Cache.
The communication resource includes between the node 210 for block chain network, between block chain network 200 and business division
The various links of communication.
Data Layer 204 encapsulates the various data structures for realizing account book, including the area realized with the file in file system
Block chain, the slip condition database and existence proof (such as the Hash tree traded in block) of key assignments type.
Network layer 203 encapsulates point-to-point (P2P, Point to Point) network protocol, data dissemination mechanism and data
The function of authentication mechanism, access authentication mechanism and business division Identity Management.
Wherein, P2P network protocol realizes the communication between 200 interior joint 210 of block chain network, and data dissemination mechanism guarantees
Propagation of the transaction in block chain network 200, data authentication mechanism are used for based on encryption method (such as digital certificate, number
Word signature, public private key-pair) realize the reliability that data are transmitted between node 210;Access authentication mechanism is used for according to actual industry
Business scene authenticates the identity for the business division that block chain network 200 is added, and business division is assigned when certification passes through
Access the permission of block chain network 200;Business division Identity Management is used to store the business master for allowing to access block chain network 200
The identity and permission (such as the type for the transaction that can be initiated) of body.
Common recognition layer 202 encapsulates the node 210 in block chain network 200 and (knows together to the mechanism of block compliance
Mechanism), the function of trade management and account book management.
Common recognition mechanism includes the common recognition algorithm such as POS, POW and DPOS, supports the pluggable of common recognition algorithm.
Trade management verifies the body of business division for verifying the digital signature carried in the transaction that node 210 receives
Part information, and judged to confirm whether it there is permission to be traded and (read phase from business subject identity management according to identity information
Close information);For obtaining the business division of authorization of access block chain network 200, possess the number that authentication center issues
Word certificate, business division signs to the transaction of submission using the private key in oneself digital certificate, to state oneself
Legal identity.
Account book management: for safeguarding block chain and account book database.For obtaining the block of common recognition, it is appended to block chain
Tail portion;The transaction in the block for obtaining common recognition is executed, the key-value pair in slip condition database is updated when transaction includes updating operation,
When transaction include inquiry operation when inquiry account book database in key-value pair and to business division return query result.Support reconciliation
The inquiry operation of a variety of dimensions of database, comprising: block is inquired according to block sequence number (such as cryptographic Hash of transaction);Root
Block is inquired according to block cryptographic Hash;Block is inquired according to transaction sequence number;It is inquired and is traded according to transaction sequence number;According to business master
The account data of account (sequence number) inquiry business main body of body;The block chain in channel is inquired according to tunnel name.
Application layer 201 encapsulates the various businesses that block chain network can be realized, tracing to the source, depositing card and verifying including transaction
Deng.
Illustrate the exemplary structure for realizing the node of the block chain network of the embodiment of the present invention below, it is possible to understand that ground, area
The hardware configuration of any type of node in block chain network 200 can be implemented according to hardware configuration described below.
It is the structural schematic diagram for the node 210 that the embodiment of the present invention is provided in block chain network 200, Fig. 3 referring to Fig. 3, Fig. 3
Shown in node 210 include: at least one processor 2110, memory 2140 and at least one network interface 2120.Node 210
In various components be coupled by bus system 2130.It is understood that bus system 2130 for realizing these components it
Between connection communication.Bus system 2130 further includes power bus, control bus and status signal in addition to including data/address bus
Bus.But for the sake of clear explanation, various buses are all designated as bus system 2130 in Fig. 3.
Processor 2110 can be a kind of IC chip, the processing capacity with signal, such as general processor, number
Word signal processor (DSP, Digital Signal Processor) either other programmable logic device, discrete gate or
Transistor logic, discrete hardware components etc., wherein general processor can be microprocessor or any conventional processing
Device etc..
Memory 2140 can be it is removable, it is non-removable or combinations thereof.Illustrative hardware device includes solid-state
Memory, hard disk drive, CD drive etc..Memory 2140 optionally includes geographically far from processor 2110
One or more storage equipment.
Memory 2140 includes volatile memory or nonvolatile memory, may also comprise volatile and non-volatile and deposits
Both reservoirs.Nonvolatile memory can be read-only memory (ROM, Read Only Me mory), and volatile memory can
To be random access memory (RAM, Random Access Memor y).2140 purport of memory of description of the embodiment of the present invention
In the memory including any suitable type.
In some embodiments, memory 2140 can storing data to support various operations, the example packet of these data
Include program, module and data structure or its subset or superset, below exemplary illustration.
Operating system 2141, including for handling various basic system services and executing the system journey of hardware dependent tasks
Sequence, such as ccf layer, core library layer, driving layer etc., for realizing various basic businesses and the hardware based task of processing;
Network communication module 2142, by being reached based on other via one or more (wired or wireless) network interfaces 2120
Equipment is calculated, illustrative network interface 2120 includes: bluetooth, Wireless Fidelity (W iFi) and universal serial bus
(USB, Universal Serial Bus) etc.;
In some embodiments, the node of block chain network provided in an embodiment of the present invention can be real using software mode
Existing, Fig. 3 shows the software module 2143 being stored in memory 2140, can be the software of the forms such as program and plug-in unit,
Comprise the following modules: model obtains module 21431, model transaction modules 21432, common recognition module 21433 and updates storage module
21434, these modules are in logic, therefore to can be combined arbitrarily according to the function of being realized or further split.
The function of modules will be described hereinafter.
Below in conjunction with the exemplary application and implementation of block chain network provided in an embodiment of the present invention, illustrate of the invention real
The machine learning model processing method based on block chain network of example offer is provided.
Referring to fig. 4, Fig. 4 is the machine learning model processing method provided in an embodiment of the present invention based on block chain network
One optional flow diagram, the step of showing in conjunction with Fig. 4, are illustrated.
In step 401, block chain network obtains the machine learning model that party in request's node is sent, by the machine learning
Model is stored.
Here, party in request's node is present in inside block chain network in the form of client node.Party in request's node-home
In business division, for business division, it is necessary first to block chain network is accessed, it is different according to the type of block chain network,
The access situation of business division is also different, such as in the case where block chain network is publicly-owned chain, business division can not be verified,
It is directly accessed block chain network;The case where block chain network is alliance's chain, settable access conditions is only full in business division
When sufficient access conditions, just business division is allowed to access, become the client node in block chain network, for example, settable access
The business division that condition is registion time 3 years or more.
Business division access block chain network is formed by other nodes of node and block chain network in initialization, to
Authentication center makes requests, so that the root certificate and digital certificate that same authenticated center is issued are configured, in this way, block chain network
In node to other nodes send information when, enclose digital certificate in the information, believe in digital certificate including the use of identity
The digital signature of signature is ceased, the node as recipient is after receiving information, the public key in root certificate held according to itself
Digital signature is decrypted, is held to verify the user whether public key in digital certificate is stated by digital certificate, is protected
Hinder the confidence level of information.
Party in request's node initiates to submit the transaction of machine learning model, transaction carry the digital certificate of party in request's node with
And party in request's node is for the digital signature (different from the above-mentioned digital signature using identity information signature) of transaction, block link network
When receiving transaction in other nodes in network, the digital certificate and digital signature carry to transaction is verified, digital certificate
And after digital signature authentication success, the identity of party in request's node is verified, whether confirmation party in request's node is with trading privilege, number
The verifying judgement failure of any one of word certificate, digital signature and Authority Verification all will lead to Fail Transaction.After being proved to be successful,
The node signs the digital signature of oneself, and digital signature and the digital certificate of itself are carried in transaction, continues in block link network
Transaction is broadcasted in network.After node in block chain network with ranking function receives the transaction being proved to be successful, transaction is filled
Into new block, and it is broadcast in block chain network and the node of common recognition service is provided.Offer in block chain network, which is known together, to be taken
The node of business carries out common recognition process to new block to reach an agreement, and new block is appended to block chain by the node for providing account book function
Tail portion, and execute the transaction in new block: i.e. for the transaction for submitting machine learning model, adding machine in account book database
The corresponding key-value pair of device learning model.It is worth noting that the account book in the embodiment of the present invention may include block chain itself with
And the slip condition database synchronous with block chain, and the embodiment of the present invention to the common recognition mechanism being related to without limitation.
In some embodiments, between arbitrary steps, the machine learning model processing method further include: the block
Chain network obtains the demand information that party in request's node is sent, and the demand information is corresponding with the machine learning model;It will
The demand information is stored.
Other than machine learning model itself, party in request's node can also be to block chain network distribution of machine learning model pair
The demand information answered, the demand information are the training requirement of machine learning model.For example, demand information instruction ownership Mr. Yu's class
The node of specific transactions main body (such as Internet enterprises) is updated machine learning model, alternatively, demand information instruction possesses
The node of the specific training data of certain class (such as communication software chat data) is updated machine learning model.Similarly, logical
After crossing the operations such as broadcast, verifying and common recognition, the successful demand information of block chain network storage common recognition, wherein can be by demand information
It is stored in a block with corresponding machine learning model, convenient for checking.After completing storage, the business master of participant node is operated
Body can judge whether participant node meets demand, to judge whether to participate in machine learning mould by obtaining demand information
The update of type.By way of above-mentioned storage demand information, the validity for updating machine learning model is improved, reduction is not inconsistent
The probability that the participant node of conjunction demand is updated machine learning model.
In some embodiments, it can also realize and above-mentioned deposit the machine learning model in this way
Storage: the block chain network includes for the unified block chain for storing different machine learning models, by the machine learning mould
Type is stored into the block chain of the block chain network;Alternatively, the block chain network include it is multiple with it is different types of
The corresponding channel of machine learning model, by machine learning model storage into the block chain in corresponding types channel.
The embodiment of the invention provides two kinds of memory mechanisms, and in a kind of memory mechanism, block chain network includes for uniting
The block chain of the different machine learning model of one storage, when getting the machine learning model of party in request's node transmission, by this
Machine learning model is stored to the block chain.In another memory mechanism, multiple channels are set in block chain network, each
Channel corresponds to a type of machine learning model, also, includes, i.e. submission mould related to machine learning model in each channel
Type and the node for participating in training.Block chain network get party in request's node transmission machine learning model after, by broadcast,
Verifying and common recognition, when knowing together successfully, by the machine learning model store into block chain network with the machine learning model
In the block chain in corresponding types channel, so that participant node in the channel carries out inquiry and more to the machine learning model
Newly.It is worth noting that participant node and party in request's node can be the shared member node in multiple channels, node is depended on
The training of which machine learning model is participated in, i.e. a participant node may be located at channel A and channel B simultaneously.One participation
Fang Jiedian exits the training of a machine learning model, that is, when exiting a channel, may still participate in the machine in other channels
The training of learning model, and updated machine learning model is sent to block chain network.It is deposited by the above-mentioned means, improving
The flexibility for storing up machine learning model, can determine the memory mechanism of block chain network according to practical application scene.
In step 402, the block chain network in response to participant node send to the machine learning model
The machine learning model is sent to the participant node by inquiry request.
Here, participant node is equally the client node inside block chain network, and participant node can be to block chain
Network sends the inquiry request to machine learning model, and block chain network broadcasts the inquiry request, verified and known together
Operation.When block chain network knows together successfully to inquiry request, block chain network will be stored in block chain according to the inquiry request
It is local that the machine learning model of network is sent to participant node.
In step 403, the participant node carries out more the machine learning model according to itself training data
Newly.
Here, the update operation that participant node executes machine learning model is unrelated with block chain network, to realize
Protection to self training data.Different, the process that participant node is updated it according to the type of machine learning model
Also different, for example, participant node can be according to training data to neural network when machine learning model is neural network model
Neural network structure in model optimizes, while carrying out to the connection weight parameter between layer each in neural network structure excellent
Change, details are not described herein again.
In step 404, the block chain network obtains the updated engineering that the participant node is sent
Model is practised, is known together to the updated machine learning model.
After participant node is completed to the update of machine learning model, participant node can initiate to mention to block chain network
Hand over the transaction of updated machine learning model, block chain network equally the operation such as broadcasts, verified and known together to the transaction.
In step 405, when knowing together successfully, the block chain network carries out the updated machine learning model
Storage.
When block chain network knows together successfully to the transaction, block chain network is according to the transaction, by updated engineering
It practises model to store into block chain network, so that the node in block chain network can check updated machine learning model.
In some embodiments, it can realize in this way above-mentioned to the updated machine learning model
Know together: the block chain network broadcasts the updated machine learning model in the block chain network,
So that the node in the block chain network fills the updated machine learning model to more new block, and to it is described more
New block carries out consistency checking.
Updated machine learning model is broadcasted the primary node into block chain network by participant node, is carried simultaneously
Digital signature and digital certificate.Machine learning model, the number of primary node upon a reception of an updated in block chain network is signed
After name and digital certificate, digital signature and digital certificate is verified, and after being proved to be successful, sign the digital signature of itself
And digital certificate, and continue to broadcast updated machine learning model in block chain network.
Node in block chain network with ranking function obtains updated machine learning model, is confirming corresponding number
On the basis of word signature and digital certificate have verified that successfully, updated machine learning model is filled into new block, is
Convenient for distinguishing, block that filling obtains is named as more new block, and more new block is broadcast to being total in block chain network
Know service node.Common recognition service node in block chain network carries out consistency checking to more new block, to reach an agreement.
It can realize above-mentioned by the updated machine learning model, be stored to the block in this way
Chain network: the more new block is appended to the tail portion of block chain.
When consistency checking success, the node that account book function is provided in block chain network more new block will be appended to block
The tail portion of chain, the block chain are present in block chain network.Updated machine learning model is realized through the above way
Cochain storage.
Implemented by above-mentioned example of the inventive embodiments for Fig. 4 it is found that the embodiment of the present invention makees block chain network
For the shared platform of machine learning model, while the process for machine learning model being updated according to training data, transfer to participant
Node is carried out in intra-node, to improve the safety for the training data that participant node is held, efficiently avoids instructing
Practice leaking data, while improving the treatment effeciency to machine learning model.
It in some embodiments, is the engineering provided in an embodiment of the present invention based on block chain network referring to Fig. 5, Fig. 5
Another the optional flow diagram for practising model treatment method, can also be in step 501 before step 402 based on Fig. 4
In, the block chain network determines authorized participant node.
Block chain network can square node according to demand instruction, participant node is authorized, and in this step
It determines authorized participant node, also determines that it is not limited in the embodiment of the present invention according to other modes certainly.
In some embodiments, before step 501, further includes: the block chain network obtains party in request's node
The information for the participant node that the needs of transmission authorize, the information of the participant node include identity information, the access area
The validity period of block chain network and the machine learning model that can be trained;When by party in request's node deployment to the block chain
The intelligent contract of network, when determining that the information of the participant node meets authorising conditional, by the participant for needing to authorize
The information of node is stored.
In embodiments of the present invention, block chain network can the instruction of square node according to demand participant node is awarded
Power.Specifically, the information for the participant node that the needs that block chain network obtains that party in request's node is sent authorize, the participant section
The information of point includes identity information, the validity period of the access block chain network and the machine learning model that can train, identity
The information such as organization of the business division of participant node-home and unit property etc., validity period refer to participant node in area
The machine learning model that can be trained can be by the mark of machine learning model, such as name there are duration in future in block chain network
Title or sequence number etc., are indicated in the information of participant node.Wherein, validity period is not configured in the information of participant node
When, participant node future is defaulted always present in block chain network;When the machine learning model that can be trained is not configured,
Default participant node can be trained the machine learning model of all kinds.It is worth noting that above-mentioned nodal information
Merely illustrative, in practical application scene, nodal information may include more and less content.
When block chain network by party in request's node deployment to block chain network node intelligent contract, determine participant
When the information of node meets authorising conditional, the information of the participant node authorized will be needed to fill to new block, and in new block
When knowing together successfully, new block is appended in block chain, in the case where existence database, while more new state data
Library realizes the cochain of the information of participant node through the above way, improve the information of participant node the property looked into and
Trackability.
In some embodiments, before step 402, further includes: the block chain network is sent out in response to participant node
Before the inquiry request to the machine learning model sent, the digital signature of the inquiry request is verified;When being verified,
In the block chain network in the information of authorized participant node, inquiry sends the participant section of the inquiry request
The information of point;When inquiring the information for sending the participant node of the inquiry request, determines and respond the inquiry request.
Participant node initiates the transaction including inquiry request to block chain network, which carries digital signature, the number
Word signature is obtained after encrypting to the abstract of the transaction.The node of block chain network is being got including inquiry request
When transaction, corresponding digital signature is verified, specifically the transaction itself is handled by node to obtain the first abstract, while to friendship
Portable digital signature is decrypted to obtain the second abstract, when the first abstract is identical as the second abstract, determines digital signature
It is verified.When being verified, in the information of the authorized participant node of storage, inquiry is sent block chain network
The information of the participant node of inquiry request determines response when inquiring the information for sending the participant node of inquiry request
Inquiry request.It is preposition to judge whether to respond inquiry request by way of the information of above-mentioned inquiry participant node, it saves
The process resource of block chain network.
In some embodiments, described to be closed when by the intelligence of party in request's node deployment to the block chain network
About, when determining that the information of the participant node meets authorising conditional, by the information of participant node for needing to authorize into
Before row storage, further includes: the block chain network obtains the intelligence including authorising conditional that party in request's node is sent and closes
About, the authorising conditional is corresponding with the machine learning model;The authorising conditional is disposed.
The authorising conditional authorized for judging whether the information to participant node, can be deployed to block link network in advance
In network.Specifically, party in request's node sends deployment transaction to block chain network, and deployment transaction will be for that will include authorising conditional
Intelligent contract is deployed to the node in block chain network, and block chain network is broadcasted, verified and known together to deployment transaction
Operation, when knowing together successfully, block chain network is traded according to the deployment, which is deployed to block chain
Node in network.In this way, party in request's node can initiate to be directed to participant in subsequent need side's node authorization participant node
The calling of the information of node is traded, and the intelligence contract disposed is called in transaction, so that block chain network is awarded to meeting
The information of the participant node of power condition carries out uplink operation.
In step 502, the block chain network is generated awards correspondingly with authorized each participant node
Weighted code, and the authorization code is sent to the corresponding participant node, the authorization code is looked into for the participant node
Ask the machine learning model.
For each authorized participant node, block chain network generates a unique authorization code, and by the authorization
Code is sent to the participant node, and the authorization code for guaranteeing that different participant node possesses is inconsistent, consequently facilitating identification identity,
In, authorization code is used to inquire the voucher of machine learning model in block chain network as participant node.It is worth noting that
Step 501 is merely illustrative to execution sequence of the step 502 in Fig. 5, according to the difference of practical application scene, can more hold
Row sequence is executed, such as step 501 to step 502 can execute before step 401.
In Fig. 5, the step 402 of Fig. 4 can be realized by step 503 to step 505, will be said in conjunction with each step
It is bright.
In step 503, the block chain network obtains the inquiry request and authorization code that the participant node is sent.
For getting the participant node of authorization code, the inquiry of machine learning model is asked to the transmission of block chain network
It asks, carries digital signature and digital certificate, while also sending authorization code.Block chain network is to the corresponding digital certificate of inquiry request
And digital signature is verified, and on the basis of digital certificate and digital signature are proved to be successful, is tested the authorization code
Card.
In step 504, the block chain network carries out purview certification according to the authorization code.
Block chain network judges whether participant node has the permission of inquiry machine learning model according to authorization code.
In some embodiments, the above-mentioned block chain network can be realized according to the authorization in this way
Code carries out purview certification: the block chain network updates secondary according to history of the participant node to the machine learning model
Number, model promotion degree, contribution data amount and historical query number, generate the authorization parameter of the authorization code;When the authorization is joined
When number meets the Parameter Conditions of setting, defines the competence and authenticate success.
In embodiments of the present invention, participant node corresponding for the authorization code, block chain network can obtain the participation
Fang Jiedian is to the history update times of machine learning model, model promotion degree, contribution data amount and historical query number, wherein
After model promotion degree refers to that participant node is updated machine learning model, carried out according to updated machine learning model
The accuracy rate promotion degree of data processing is mainly used for measuring the training data of participant node for the effective of machine learning model
Property, for the ease of quantization, model promotion degree can be embodied in after the updated machine learning model of application, be mentioned to specific business
Liter degree, such as to the promotion degree for launching success rate or clicking rate etc.;Contribution data amount can refer to participant node to machine learning mould
The total quantity of training data used when type is updated also can refer to the total quantity of training data used in participant node and update time
Ratio between number, the i.e. par of the training data used in each update.Above-mentioned history update times, model mention
Liter degree, contribution data amount and historical query number can be stored in the account book of block chain network, consequently facilitating obtaining.
Then, pass through function G ' (S)=f ' (history update times, model promotion degree, contribution data amount, historical query time
Number) authorized parameter G ' (S), wherein between history update times, model promotion degree and contribution data amount and G ' (S) at
Positive correlation is negatively correlated relationship between historical query number and G ' (S), by weighted summation, authorized ginseng
Number G ' (S).When authorization parameter meets the Parameter Conditions of setting, defines the competence and authenticate success, the corresponding participant section of authorization code
Point has the permission of inquiry machine learning model;When authorization parameter is unsatisfactory for Parameter Conditions, define the competence authentification failure,
In, Parameter Conditions such as authorization parameter is greater than zero.It is above-mentioned in such a way that authorization parameter carries out purview certification, be to participate in actually
Fang Jiedian is provided with incentive mechanism, i.e. participant node only plays an active part in the update to machine learning model, could obtain more
More inquiry (use) numbers, so that more participant nodes is encouraged to participate in the update of machine learning model, elevator
The more new effects of device learning model.
In step 505, when purview certification success, the block chain network, will be described in response to the inquiry request
Machine learning model is sent to the participant node.
When purview certification success, block chain network is sent to ginseng according to inquiry request, by corresponding machine learning model
With Fang Jiedian;When permission authentification failure, block chain network does not respond inquiry request, i.e. not distribution of machine learning model.
In some embodiments, after step 505, further includes: the block chain network obtains the participant node
The participation information of transmission, wherein the update that the participation information carries out the machine learning model with the participant node
Operation is related;The participation information is stored.
Participant node is also transmittable to grasp with update other than sending updated machine learning model to block chain network
Make relevant participation information, for example, participation information may include the contribution data amount that participant node this time updates, Yi Jiji
The information such as the update degree of device learning model.Block chain network obtains the participation information that participant node is sent, and believes participation
Cease the operation such as broadcasted, verified and known together.Wherein, party in request's node can be submitted in conjunction with above content, block chain network
The updated machine learning model and participation information that demand information, participant node are submitted are configured to a block, and right
The block of building is known together.
In successful situation of knowing together, block chain network stores participation information, so that determining participant node
Authorization parameter or when other situations, block chain network can obtain relevant information from participation information, such as contribution data amount.
In some embodiments, between arbitrary steps, the machine learning model processing method further include: the block
Chain network determines the model income of the machine learning model;The machine learning model is gone through according to the participant node
History update times, model promotion degree and contribution data amount determine the contribution parameters of the participant node;According to each participation
The contribution parameters of Fang Jiedian determine corresponding model income, by the model distribution of income to each participant node.
Block chain network can determine the model income of machine learning model, and be allocated to model income, model income
It can also be able to be the income of other forms, the present invention for machine learning model is applied to economic well-being of workers and staff obtained from production field
Embodiment does not limit this.Specifically, function G (S)=f (history update times, model promotion degree, contribution data can be passed through
Amount) it obtains contribution parameters G (S), which embodies participant node to the percentage contribution of final machine learning model,
Wherein, equal positive correlation between history update times, model promotion degree and contribution data amount and G (S), passes through weighted sum
Mode, obtain contribution parameters G (S).In addition to this, when determining contribution parameters, it is also contemplated that the history of participant node
The factor of inquiry times, that is, in embodiments of the present invention, can be using above-mentioned authorization parameter as contribution parameters.It is worth explanation
It is that above-mentioned history update times, model promotion degree and contribution data amount can be filled into block by node, and by knowing together
After be stored in block chain, on this basis, if being provided with channel in block chain network, by history update times, model
Promotion degree and contribution data amount are stored into the block chain of respective channel, in case inquiry.
According to the difference of percentage contribution, block chain network is by model distribution of income to each participant node.Specifically, it distributes
To participant node SiModel income benefit (Si)=machine learning model model income * G (Si)/(G(S1)+G(S2)
+……+G(Sn)), wherein n is the sum for participating in the participant node that machine learning model updates.It is worth noting that block
Chain network can carry out the distribution of model income by executing corresponding intelligent contract, and specifically, block chain network obtains party in request
The deployment transaction comprising intelligent contract that node is sent, and intelligent contract is deployed in the node of block chain network, the intelligence
Contract is the intelligent contract for distribution model income.After the completion of deployment, block chain network obtains what party in request's node was sent
Calling transaction to intelligent contract, to determine the model income of machine learning model according to intelligent contract, determine participant section
The contribution parameters of point determine the corresponding model income of each participant node according to contribution parameters, and by model distribution of income to each
Participant node.The mode of above-mentioned distribution model income, is another incentive mechanism of setting actually, encourages to join by income
Machine learning model is updated with square node, thus the more new effects of hoisting machine learning model.In some embodiments,
It can realize in this way above-mentioned by the model distribution of income to each participant node: the block link network
Network stores the distribution information of the model income of each participant node, and sends accordingly to each participant node
Mark of the model income in the block chain network;Wherein, the model income includes at least one of: accessing the area
The validity period of block chain network, the free memory in the block chain network are inquired or are deposited in the block chain network
Store up the frequency of machine learning model.
In embodiments of the present invention, block chain network is after determining the corresponding model income of each participant node, can will be each
The distribution information of the model income of participant node is stored, and sends corresponding model income in block to each participant node
Mark in chain network so that participant node according to mark in block chain network interrogation model income, wherein can be to distribution
Information carries out hashing operation and is identified, and mark can also be configured according to other modes certainly.Model income is received in addition to economy
Beneficial outer, may also be configured to include at least one of: participant node accesses the validity period of block chain network, such as 10 days, in area
Free memory in block chain network, such as 1 gigabytes inquire or store the frequency of machine learning model in block chain network
Rate, such as most 1 day 3 times.Above-mentioned by way of the distribution information of block chain network storage model income, model is improved
The trackability of income.
Implemented by above-mentioned example of the inventive embodiments for Fig. 5 it is found that the embodiment of the present invention is carried out by authorization code
Purview certification only enables the participant node with permission inquire to obtain machine learning model, improves machine learning model
Query safe.
It in some embodiments, is machine learning model processing method provided in an embodiment of the present invention referring to Fig. 6, Fig. 6
Another optional flow diagram, can also in step 601 before step 402 based on Fig. 4, the block chain network
Determine authorized participant node.
In embodiments of the present invention, settable encryption mechanism ensures the data safety of machine learning model itself.Specifically,
The instruction of block chain network square node according to demand, determines authorized participant node, certainly, authorized participant node
It can also be determined according to other modes.
In step 602, the block chain network generates close correspondingly with authorized each participant node
Key pair, and the key pair is sent to the corresponding participant node.
In Fig. 6, the machine learning model from party in request's node that block chain network is got is added using public key
It is close, therefore block chain network is for authorized each participant node, by rivest, shamir, adelman, generate include the public key and with
The key pair of the matched unique private of the public key, and key pair is sent to corresponding participant node, it is in addition to this, also independent
A set of key pair is generated, party in request's node is sent to.For example, public key A, authorized participant node include node 1
With node 2, then the key pair including public key A and matched private key B1 is generated, which is sent to node 1;Generation includes
The key pair is sent to node 2 by the key pair of public key A and matched private key B2.It is worth noting that above-mentioned key pair
It can be generated by participant node.It is worth noting that execution sequence merely illustrative, root of the step 601 to step 602 in Fig. 6
It according to the difference of practical application scene, can be executed using different execution sequences, such as step 601 to step 602 can be
It is executed before step 401.
In Fig. 6, the step 403 shown in Fig. 4 be may be updated as:
In step 403, the participant node sends out party in request's node according to the private key for the cipher key pair held
The machine learning model for the encryption sent is decrypted, and according to the training data of itself, carries out to the machine learning model
It updates, and the updated machine learning model is encrypted according to the public key for the cipher key pair held.
Participant node is after inquiring the machine learning model using public key encryption, the cipher key pair held according to itself
Private key, machine learning model is decrypted.After completing decryption, participant node is according to itself training data, to machine
Learning model is updated, and is encrypted using the public key of cipher key pair to updated machine learning model.Due to being awarded
The public key that each participant node of power is held is all the same, therefore updated machine learning model is being stored to block chain network,
It can be by other participant node successful decryptions.
On this basis, when party in request's node submits demand information to block chain network, due to the usual nothing of demand information
It must maintain secrecy, therefore party in request's node can submit original, the i.e. demand information of unencryption to block chain network.In addition, for
Whether the participation information that participant node is submitted to block chain network can be arranged according to practical application scene and be carried out with public key to it
Encryption.For the case where setting encrypts participation information with public key, the block constructed in block chain network can be wrapped
Include: 1) demand information that party in request's node is submitted is issued in plain text using open;2) machine learning model specifically includes model
Structural information and parameter information are issued using encryption;3) participation information that participant node is submitted is issued using encryption.
Implemented by above-mentioned example of the inventive embodiments for Fig. 6 it is found that the embodiment of the present invention passes through public key to machine
Learning model is encrypted, guarantee the machine learning model of block chain network storage will not held the node of corresponding private key at
Function decryption, improves the safety of storage machine learning model.
In the following, will illustrate exemplary application of the embodiment of the present invention in actual application scenarios.
As shown in fig. 7, another optional process the embodiment of the invention provides machine learning model processing method is shown
It is intended to, in order to make it easy to understand, being illustrated in the form of number to Fig. 7 hereinafter.
1) party in request's node initiates request, and demand information is stored to block chain network.
Party in request's node is the node inside block chain network, when needing to carry out model training, is sent out to block chain network
Request is played, demand information is sent to block chain network, demand information is stored by block chain network, wherein demand letter
Breath may include the correlative detail of machine learning model and the demand to training data.
2) party in request's node issues initial machine learning model, and machine learning model encryption is stored to block link network
Network.
Party in request's node encrypts initial machine learning model using public key, and by encrypted machine learning mould
Type is sent to block chain network, is stored by block chain network to it.
3) participant node is according to itself training data, it is determined whether participates in the training to machine learning model.
Participant node is inquired from block chain network and obtains demand information, and judge itself training data whether with need
Seek information matches.For example, the training data of demand information instruction demand is communication software chat data, and participant node itself
Training data be game data, then participant node determines that the training data of itself and demand information mismatch, and is not involved in pair
The training of machine learning model.
4) participant node is inquired from block chain network and obtains encrypted machine learning model, according to the private key held
Encrypted machine learning model is decrypted;Machine learning model is updated according to the training data of itself.
For encrypting public key used, block chain network or participant node generate multiple unique privates matched with the public key
Key constitutes multiple key pairs, and multiple key pairs is sent to the different participant nodes with permission one by one.Participant section
Point is determining that inquiry obtains encrypted machine learning mould from block chain network when participating in the training to machine learning model
Type is decrypted according to the machine learning model after the private key pair encryption held, then, according to the training data of itself to machine
Learning model is updated.
5) participant nodes records participation information relevant to operation is updated, by updated machine learning model and ginseng
It stores with information encryption to block chain network.
Participant nodes records update the participation information generated when machine learning model, for example, participation information may include ginseng
With the contribution data amount of square node, and the update degree to machine learning model.It completes after updating, participant node will update
Machine learning model and participation information afterwards is encrypted by private key, and is stored by common recognition mechanism to block link network
Network.In conjunction with above content, the block in block chain may include following information: encrypt the demand information of publication, the machine of encryption publication
Device learning model and the participation information of encryption publication.In the follow-up process, participant node can be determined whether to continue to participate in pair
The training of machine learning model.
Implemented by above-mentioned example of the inventive embodiments for Fig. 7 it is found that the embodiment of the present invention makees block chain network
For the network platform for sharing machine learning model, so that the node having permission is able to carry out storage to machine learning model, more
New and inquiry, meanwhile, the process of machine learning model will be updated according to training data, transfer to participant node intra-node into
Row efficiently avoids training data leakage to improve the safety for the training data that participant node is held.
It continues with and illustrates that the node of block chain network provided in an embodiment of the present invention is embodied as the exemplary of software module
Structure, in some embodiments, as shown in figure 3, the software module 2143 for being stored in memory 2140 may include: that model obtains
Module 21431 stores the machine learning model to block chain for obtaining the machine learning model of party in request's node transmission
Network;Model transaction modules 21432 ask the inquiry of the machine learning model for what is sent in response to participant node
It asks, the machine learning model is sent to the participant node, so that training number of the participant node according to itself
According to being updated to the machine learning model;Common recognition module 21433, the update sent for obtaining the participant node
The machine learning model afterwards knows together to the updated machine learning model;Module 21434 is updated storage, is used
In when knowing together successfully, the updated machine learning model is stored to the block chain network.
In some embodiments, software module 2143 further include: first node determining module, for determining authorized ginseng
With Fang Jiedian;Authorization module, each one-to-one authorization code of participant node for generating and being authorized to, and will be described
Authorization code is sent to the corresponding participant node, and the authorization code is for machine learning described in the participant querying node
Model.
In some embodiments, model transaction modules 21432, are also used to: obtaining the authorization that the participant node is sent
Code;Purview certification is carried out according to the authorization code;When purview certification success, the machine learning model is sent to the ginseng
With Fang Jiedian.
In some embodiments, described that purview certification is carried out according to the authorization code, comprising: according to the participant node
To the history update times of the machine learning model, model promotion degree, contribution data amount and historical query number, described in generation
The authorization parameter of authorization code;When the authorization parameter meets the Parameter Conditions of setting, defines the competence and authenticate success.
In some embodiments, software module 2143 further include: data obtaining module, for obtaining party in request's node
The information for the participant node that the needs of transmission authorize, the information of the participant node include identity information, the access area
The validity period of block chain network and the machine learning model that can be trained;Information storage module is saved for working as by the party in request
Point is deployed to the intelligent contract of the block chain network, when determining that the information of the participant node meets authorising conditional, by institute
It states and the information of the participant node authorized is needed to store into the block chain network.
In some embodiments, software module 2143 further include: authentication module, for what is sent in response to participant node
Before the inquiry request of the machine learning model, the digital signature of the inquiry request is verified;Information inquiry module is used for
When being verified, in the block chain network in the information of authorized participant node, inquiry sends the inquiry
The information of the participant node of request;Respond module, for when the letter for inquiring the participant node for sending the inquiry request
When breath, determines and respond the inquiry request.
In some embodiments, software module 2143 further include: contract obtains module, for obtaining party in request's node
The intelligent contract including authorising conditional sent, the authorising conditional are corresponding with the machine learning model;Contract deployment module,
For the authorising conditional to be deployed in the block chain network.
In some embodiments, software module 2143 further include: income determining module, for determining the machine learning mould
The model income of type;Determining module is contributed, for updating according to history of the participant node to the machine learning model
Number, model promotion degree and contribution data amount determine the contribution parameters of the participant node;Distribution module, for according to each
The contribution parameters of the participant node determine corresponding model income, by the model distribution of income to each participant section
Point.
In some embodiments, it is described include: to each participant node by the model distribution of income will be each described
The distribution information of the model income of participant node is stored into the block chain network, and is sent to each participant node
Mark of the corresponding model income in the block chain network;Wherein, the model income includes at least one of: access institute
The validity period for stating block chain network, the free memory in the block chain network are inquired in the block chain network
Or the frequency of storage machine learning model.
In some embodiments, software module 2143 further include: second node determining module, for determining authorized ginseng
With Fang Jiedian;Key sending module, each one-to-one key pair of participant node for generating and being authorized to, and will
The key pair is sent to the corresponding participant node, so that private of the participant node according to the cipher key pair held
The machine learning model for the encryption that party in request's node is sent is decrypted in key, and according to the cipher key pair held
Public key the updated machine learning model is encrypted, wherein the cipher key pair of the different participant nodes
Private key is different.
In some embodiments, software module 2143 further include: participation information common recognition module, for obtaining the participant
The participation information that node is sent, knows together to the participation information, wherein the participation information and the participant node pair
The update that the machine learning model carries out operates related;Participation information memory module is used for when knowing together successfully, by the ginseng
It stores with information to the block chain network.
In some embodiments, software module 2143 further include: demand information obtains module, for obtaining the party in request
The demand information that node is sent, the demand information are corresponding with the machine learning model;Demand information memory module, being used for will
The demand information is stored to the block chain network.
In some embodiments, common recognition module 21433 is also used to: by the updated machine learning model in the area
Broadcasted in block chain network so that the node in the block chain network by the updated machine learning model fill to
More new block, and consistency checking is carried out to the more new block;
It updates storage module 21434 to be also used to: the more new block is appended to the tail portion of block chain.
In some embodiments, the model obtains module 21431 and is also used to: the block chain network includes for unified
The block chain for storing different machine learning models, by machine learning model storage to the area of the block chain network
In block chain;Alternatively, the block chain network includes multiple channels corresponding with different types of machine learning model, by the machine
Device learning model is stored into the block chain in corresponding types channel.
The embodiment of the present invention provides a kind of storage medium for being stored with executable instruction, wherein it is stored with executable instruction,
When executable instruction is executed by processor, processor will be caused to execute method provided in an embodiment of the present invention, for example, such as Fig. 4,
5, the machine learning model processing method shown in 6 or 7.
In some embodiments, storage medium can be FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface and deposit
The memories such as reservoir, CD or CD-ROM;Be also possible to include one of above-mentioned memory or any combination various equipment.
In some embodiments, executable instruction can use program, software, software module, the form of script or code,
By any form of programming language (including compiling or interpretative code, or declaratively or process programming language) write, and its
It can be disposed by arbitrary form, including be deployed as independent program or be deployed as module, component, subroutine or be suitble to
Calculate other units used in environment.
As an example, executable instruction can with but not necessarily correspond to the file in file system, can be stored in
A part of the file of other programs or data is saved, for example, being stored in hypertext markup language (HTML, Hyper Text
Markup Language) in one or more scripts in document, it is stored in the single file for being exclusively used in discussed program
In, alternatively, being stored in multiple coordinated files (for example, the file for storing one or more modules, subprogram or code section).
As an example, executable instruction can be deployed as executing in a calculating equipment, or it is being located at one place
Multiple calculating equipment on execute, or, be distributed in multiple places and by multiple calculating equipment of interconnection of telecommunication network
Upper execution.
In conclusion the embodiment of the present invention is using block chain network as the network platform of shared machine learning model, so that
The node having permission is able to carry out storage, update and inquiry operation to machine learning model, meanwhile, it will be according to training data more
The process of new engine learning model transfers to participant node to carry out in intra-node, to improve what participant node was held
The safety of training data, in the machine learning model Training scene participated in many ways, compared to traditional transaction training data
Mode, efficiently avoid training data leakage, improve the safety of training data, while improving machine learning model
Treatment effeciency.
The above, only the embodiment of the present invention, are not intended to limit the scope of the present invention.It is all in this hair
Made any modifications, equivalent replacements, and improvements etc. within bright spirit and scope, be all contained in protection scope of the present invention it
It is interior.
Claims (15)
1. a kind of machine learning model processing method based on block chain network characterized by comprising
The machine learning model that party in request's node is sent is obtained, the machine learning model is stored to block chain network;
In response to the inquiry request to the machine learning model that participant node is sent, the machine learning model is sent
To the participant node, so that training data of the participant node according to itself, carries out the machine learning model
It updates;
The updated machine learning model that the participant node is sent is obtained, to the updated machine learning mould
Type is known together;
When knowing together successfully, the updated machine learning model is stored to the block chain network.
2. machine learning model processing method according to claim 1, which is characterized in that further include:
Determine authorized participant node;
Generate with authorized each one-to-one authorization code of participant node, and the authorization code is sent to corresponding
The participant node, the authorization code is for machine learning model described in the participant querying node.
3. machine learning model processing method according to claim 2, which is characterized in that described by the machine learning mould
Type is sent to the participant node, comprising:
Obtain the authorization code that the participant node is sent;
Purview certification is carried out according to the authorization code;
When purview certification success, the machine learning model is sent to the participant node.
4. machine learning model processing method according to claim 3, which is characterized in that it is described according to the authorization code into
Row purview certification, comprising:
History update times, model promotion degree, contribution data amount according to the participant node to the machine learning model
And historical query number, generate the authorization parameter of the authorization code;
When the authorization parameter meets the Parameter Conditions of setting, defines the competence and authenticate success.
5. machine learning model processing method according to claim 2, which is characterized in that further include:
Obtain the information for the participant node that the needs that party in request's node is sent authorize, the packet of the participant node
The machine learning model for including identity information, accessing the validity period of the block chain network and capable of training;
When the intelligent contract by party in request's node deployment to the block chain network, the letter of the participant node is determined
When breath meets authorising conditional, the information for needing the participant node authorized is stored into the block chain network.
6. machine learning model processing method according to claim 5, which is characterized in that further include:
Before the inquiry request to the machine learning model sent in response to participant node, the inquiry request is verified
Digital signature;
When being verified, in the block chain network in the information of authorized participant node, described in inquiry is sent
The information of the participant node of inquiry request;
When inquiring the information for sending the participant node of the inquiry request, determines and respond the inquiry request.
7. machine learning model processing method according to claim 5, which is characterized in that further include:
Obtain the intelligent contract including authorising conditional that party in request's node is sent, the authorising conditional and the machine learning
Model is corresponding;
The authorising conditional is deployed in the block chain network.
8. machine learning model processing method according to claim 1, which is characterized in that further include:
Determine the model income of the machine learning model;
According to the participant node to the history update times of the machine learning model, model promotion degree and contribution data
Amount, determines the contribution parameters of the participant node;
Corresponding model income is determined according to the contribution parameters of each participant node, by the model distribution of income to each institute
State participant node.
9. machine learning model processing method according to claim 8, which is characterized in that described by the model income point
It is assigned to each participant node, comprising:
The distribution information of the model income of each participant node is stored into the block chain network, and to each ginseng
Mark of the corresponding model income in the block chain network is sent with square node;
Wherein, the model income includes at least one of: the validity period of the block chain network is accessed, in the block chain
Free memory in network inquires or stores the frequency of machine learning model in the block chain network.
10. machine learning model processing method according to claim 1, which is characterized in that further include:
Determine authorized participant node;
Generate with authorized each one-to-one key pair of participant node, and the key pair is sent to corresponding
The participant node, so that
The participant node is according to the private key of the cipher key pair held, to the machine for the encryption that party in request's node is sent
Device learning model is decrypted, and is added according to the public key for the cipher key pair held to the updated machine learning model
It is close,
Wherein, the private key of the cipher key pair of the different participant nodes is different.
11. machine learning model processing method according to claim 1, which is characterized in that further include:
Obtain the participation information that the participant node is sent, wherein the participation information is with the participant node to described
The update that machine learning model carries out operates related;
The participation information is stored to the block chain network.
12. machine learning model processing method according to claim 1, which is characterized in that further include:
The demand information that party in request's node is sent is obtained, the demand information is corresponding with the machine learning model;
The demand information is stored to the block chain network.
13. according to claim 1 to 12 described in any item machine learning model processing methods, which is characterized in that
It is described to know together to the updated machine learning model, comprising:
The updated machine learning model is broadcasted in the block chain network, so that
Node in the block chain network fills the updated machine learning model to more new block, and to it is described more
New block carries out consistency checking;
It is described by the updated machine learning model, store to the block chain network, comprising:
The more new block is appended to the tail portion of block chain.
14. a kind of node of block chain network characterized by comprising
Model obtains module, and for obtaining the machine learning model of party in request's node transmission, the machine learning model is stored
To block chain network;
Model transaction modules, the inquiry request to the machine learning model for being sent in response to participant node, by institute
It states machine learning model and is sent to the participant node, so that training data of the participant node according to itself, to institute
Machine learning model is stated to be updated;
Common recognition module, the updated machine learning model sent for obtaining the participant node, to updated
The machine learning model is known together;
Module is updated storage, for when knowing together successfully, the updated machine learning model to be stored to the block chain
Network.
15. a kind of node of block chain network characterized by comprising
Memory, for storing executable instruction;
Processor when for executing the executable instruction stored in the memory, is realized described in any one of claim 1 to 13
Machine learning model processing method.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910776690.2A CN110490305B (en) | 2019-08-22 | 2019-08-22 | Machine learning model processing method based on block chain network and node |
CN201911295577.9A CN110991622B (en) | 2019-08-22 | 2019-08-22 | Machine learning model processing method based on block chain network and node |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910776690.2A CN110490305B (en) | 2019-08-22 | 2019-08-22 | Machine learning model processing method based on block chain network and node |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911295577.9A Division CN110991622B (en) | 2019-08-22 | 2019-08-22 | Machine learning model processing method based on block chain network and node |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110490305A true CN110490305A (en) | 2019-11-22 |
CN110490305B CN110490305B (en) | 2021-06-04 |
Family
ID=68552751
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911295577.9A Active CN110991622B (en) | 2019-08-22 | 2019-08-22 | Machine learning model processing method based on block chain network and node |
CN201910776690.2A Active CN110490305B (en) | 2019-08-22 | 2019-08-22 | Machine learning model processing method based on block chain network and node |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911295577.9A Active CN110991622B (en) | 2019-08-22 | 2019-08-22 | Machine learning model processing method based on block chain network and node |
Country Status (1)
Country | Link |
---|---|
CN (2) | CN110991622B (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111062044A (en) * | 2019-12-09 | 2020-04-24 | 支付宝(杭州)信息技术有限公司 | Model joint training method and device based on block chain |
CN111753984A (en) * | 2020-06-28 | 2020-10-09 | 中国银行股份有限公司 | Distributed AI training method, device and system based on block chain |
CN111859415A (en) * | 2020-06-18 | 2020-10-30 | 上海艾麒信息科技有限公司 | Neural network model encryption system and method |
CN111882074A (en) * | 2020-07-30 | 2020-11-03 | 平安科技(深圳)有限公司 | Data preprocessing system, method, computer device and readable storage medium |
CN112016911A (en) * | 2020-08-20 | 2020-12-01 | 王红根 | Information management method based on block chain digital currency finance and digital service platform |
CN112163034A (en) * | 2020-09-03 | 2021-01-01 | 中国联合网络通信集团有限公司 | Model sharing method, node and block chain system |
CN112348063A (en) * | 2020-10-27 | 2021-02-09 | 广东电网有限责任公司电力调度控制中心 | Model training method and device based on federal transfer learning in Internet of things |
CN112699414A (en) * | 2020-08-14 | 2021-04-23 | 武汉乾云软件开发中心(有限合伙) | Block chain-based machine learning data evidence storage method and system |
CN112801134A (en) * | 2020-12-30 | 2021-05-14 | 杭州趣链科技有限公司 | Gesture recognition model training and distributing method and device based on block chain and image |
CN113420017A (en) * | 2021-06-21 | 2021-09-21 | 上海特高信息技术有限公司 | Block chain application method for acquiring robot navigation algorithm training data set |
CN113767386A (en) * | 2020-01-27 | 2021-12-07 | 慧与发展有限责任合伙企业 | Security parameter merging using homomorphic encryption for group learning |
CN114037086A (en) * | 2021-10-19 | 2022-02-11 | 中国联合网络通信集团有限公司 | Block chain-based machine learning task distribution method, equipment and system |
CN114707663A (en) * | 2022-04-15 | 2022-07-05 | 中国电信股份有限公司 | Distributed machine learning method and device, electronic equipment and storage medium |
CN116566846A (en) * | 2023-07-05 | 2023-08-08 | 中国电信股份有限公司 | Model management method and system, shared node and network node |
US11748835B2 (en) | 2020-01-27 | 2023-09-05 | Hewlett Packard Enterprise Development Lp | Systems and methods for monetizing data in decentralized model building for machine learning using a blockchain |
US11966818B2 (en) | 2019-02-21 | 2024-04-23 | Hewlett Packard Enterprise Development Lp | System and method for self-healing in decentralized model building for machine learning using blockchain |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20220198304A1 (en) * | 2020-12-18 | 2022-06-23 | International Business Machines Corporation | Providing explainable machine learning model results using distributed ledgers |
CN115250485A (en) * | 2021-04-27 | 2022-10-28 | 华为技术有限公司 | Method and device for distributing models |
CN113612598B (en) * | 2021-08-02 | 2024-02-23 | 北京邮电大学 | Internet of vehicles data sharing system and method based on secret sharing and federal learning |
WO2023096637A1 (en) * | 2021-11-23 | 2023-06-01 | Visa International Service Association | Automatic model onboarding and searching-based optimization |
Citations (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170091397A1 (en) * | 2012-01-26 | 2017-03-30 | Netspective Communications Llc | Device-driven non-intermediated blockchain system over a social integrity network |
CN108234127A (en) * | 2016-12-22 | 2018-06-29 | ***通信集团公司 | A kind of Internet of Things method and device based on block chain |
CN108270780A (en) * | 2018-01-08 | 2018-07-10 | 中国电子科技集团公司第三十研究所 | A kind of heterogeneous network environment multicenter digital identity management method |
CN108323200A (en) * | 2018-01-25 | 2018-07-24 | 深圳前海达闼云端智能科技有限公司 | Data training method and device based on block chain, storage medium and block chain link points |
CN108563784A (en) * | 2018-04-26 | 2018-09-21 | 福建中锐海沃科技有限公司 | A kind of digital resource adduction relationship recording method and system |
CN108768700A (en) * | 2018-05-11 | 2018-11-06 | 北京奇虎科技有限公司 | Reward distribution method, apparatus and system in a kind of block chain |
CN109034837A (en) * | 2018-06-29 | 2018-12-18 | 山东华夏维康农牧科技有限公司 | Multi-code is traced to the source anti-fake method and system |
CN109167695A (en) * | 2018-10-26 | 2019-01-08 | 深圳前海微众银行股份有限公司 | Alliance Network construction method, equipment and readable storage medium storing program for executing based on federation's study |
CN109242636A (en) * | 2018-09-26 | 2019-01-18 | 盈盈(杭州)网络技术有限公司 | A kind of data transacting system and its implementation based on block chain |
CN109450856A (en) * | 2018-10-12 | 2019-03-08 | 西安电子科技大学 | Data-link information flow control system and method based on block chain |
CN109542856A (en) * | 2018-11-26 | 2019-03-29 | 上海基诺联生物科技有限公司 | A kind of Distributed data share system and method based on block chain |
CN109670583A (en) * | 2018-12-27 | 2019-04-23 | 浙江省公众信息产业有限公司 | Data analysing method, system and the medium of decentralization |
US10270599B2 (en) * | 2017-04-27 | 2019-04-23 | Factom, Inc. | Data reproducibility using blockchains |
CN109698822A (en) * | 2018-11-28 | 2019-04-30 | 众安信息技术服务有限公司 | Combination learning method and system based on publicly-owned block chain and encryption neural network |
CN109711408A (en) * | 2018-12-28 | 2019-05-03 | 天津智芯视界科技有限公司 | A kind of personal identification system for construction site |
CN109716346A (en) * | 2016-07-18 | 2019-05-03 | 河谷生物组学有限责任公司 | Distributed machines learning system, device and method |
CN109873866A (en) * | 2019-02-20 | 2019-06-11 | 北京邮电大学 | A kind of data sharing method based on block chain, device and electronic equipment |
US20190188787A1 (en) * | 2017-12-20 | 2019-06-20 | Accenture Global Solutions Limited | Analytics engine for multiple blockchain nodes |
CN110020774A (en) * | 2019-01-30 | 2019-07-16 | 阿里巴巴集团控股有限公司 | A kind of resource share method, system, device and electronic equipment |
Family Cites Families (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109891416A (en) * | 2016-10-27 | 2019-06-14 | 株式会社电装 | For authenticating and the system and method for authorization device |
US20190043024A1 (en) * | 2017-08-05 | 2019-02-07 | Proclus Technologies Limited | Method and System for Securing a Blockchain with Proof-of-Transactions |
CN109685511B (en) * | 2018-05-30 | 2023-06-09 | 上海分壳信息技术股份有限公司 | Block chain-based data use right transaction method |
CN108898219A (en) * | 2018-06-07 | 2018-11-27 | 广东工业大学 | A kind of neural network training method based on block chain, device and medium |
CN109493216B (en) * | 2018-09-30 | 2021-02-09 | 北京小米移动软件有限公司 | Model training method, device, system and storage medium |
CN109543726A (en) * | 2018-11-06 | 2019-03-29 | 联动优势科技有限公司 | A kind of method and device of training pattern |
CN109802940A (en) * | 2018-12-12 | 2019-05-24 | 北京众享比特科技有限公司 | Block chain data base encryption and decryption method, device, equipment and its storage medium |
CN109922155B (en) * | 2019-03-18 | 2022-03-04 | 众安信息技术服务有限公司 | Method and device for realizing intelligent agent in block chain network |
CN110034917A (en) * | 2019-04-11 | 2019-07-19 | 鸿秦(北京)科技有限公司 | A kind of alliance's chain data processing method and device based on homomorphic encryption algorithm |
-
2019
- 2019-08-22 CN CN201911295577.9A patent/CN110991622B/en active Active
- 2019-08-22 CN CN201910776690.2A patent/CN110490305B/en active Active
Patent Citations (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170091397A1 (en) * | 2012-01-26 | 2017-03-30 | Netspective Communications Llc | Device-driven non-intermediated blockchain system over a social integrity network |
CN109716346A (en) * | 2016-07-18 | 2019-05-03 | 河谷生物组学有限责任公司 | Distributed machines learning system, device and method |
CN108234127A (en) * | 2016-12-22 | 2018-06-29 | ***通信集团公司 | A kind of Internet of Things method and device based on block chain |
US10270599B2 (en) * | 2017-04-27 | 2019-04-23 | Factom, Inc. | Data reproducibility using blockchains |
US20190188787A1 (en) * | 2017-12-20 | 2019-06-20 | Accenture Global Solutions Limited | Analytics engine for multiple blockchain nodes |
CN108270780A (en) * | 2018-01-08 | 2018-07-10 | 中国电子科技集团公司第三十研究所 | A kind of heterogeneous network environment multicenter digital identity management method |
CN108323200A (en) * | 2018-01-25 | 2018-07-24 | 深圳前海达闼云端智能科技有限公司 | Data training method and device based on block chain, storage medium and block chain link points |
CN108563784A (en) * | 2018-04-26 | 2018-09-21 | 福建中锐海沃科技有限公司 | A kind of digital resource adduction relationship recording method and system |
CN108768700A (en) * | 2018-05-11 | 2018-11-06 | 北京奇虎科技有限公司 | Reward distribution method, apparatus and system in a kind of block chain |
CN109034837A (en) * | 2018-06-29 | 2018-12-18 | 山东华夏维康农牧科技有限公司 | Multi-code is traced to the source anti-fake method and system |
CN109242636A (en) * | 2018-09-26 | 2019-01-18 | 盈盈(杭州)网络技术有限公司 | A kind of data transacting system and its implementation based on block chain |
CN109450856A (en) * | 2018-10-12 | 2019-03-08 | 西安电子科技大学 | Data-link information flow control system and method based on block chain |
CN109167695A (en) * | 2018-10-26 | 2019-01-08 | 深圳前海微众银行股份有限公司 | Alliance Network construction method, equipment and readable storage medium storing program for executing based on federation's study |
CN109542856A (en) * | 2018-11-26 | 2019-03-29 | 上海基诺联生物科技有限公司 | A kind of Distributed data share system and method based on block chain |
CN109698822A (en) * | 2018-11-28 | 2019-04-30 | 众安信息技术服务有限公司 | Combination learning method and system based on publicly-owned block chain and encryption neural network |
CN109670583A (en) * | 2018-12-27 | 2019-04-23 | 浙江省公众信息产业有限公司 | Data analysing method, system and the medium of decentralization |
CN109711408A (en) * | 2018-12-28 | 2019-05-03 | 天津智芯视界科技有限公司 | A kind of personal identification system for construction site |
CN110020774A (en) * | 2019-01-30 | 2019-07-16 | 阿里巴巴集团控股有限公司 | A kind of resource share method, system, device and electronic equipment |
CN109873866A (en) * | 2019-02-20 | 2019-06-11 | 北京邮电大学 | A kind of data sharing method based on block chain, device and electronic equipment |
Non-Patent Citations (3)
Title |
---|
KIM H等: "" On-device federated learning via blockchain and its latency analysis"", 《COMPUTER SCIENCE 》 * |
QIANG YANG等: ""Federated Machine Learning: Concept and Applications"", 《ACM TRANSACTIONS ON INTELLIGENT SYSTEMS AND TECHNOLOGY》 * |
*** 等: ""基于机器学习的区块链智能合约脚本设计"", 《情报工程》 * |
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11966818B2 (en) | 2019-02-21 | 2024-04-23 | Hewlett Packard Enterprise Development Lp | System and method for self-healing in decentralized model building for machine learning using blockchain |
CN111062044B (en) * | 2019-12-09 | 2021-03-23 | 支付宝(杭州)信息技术有限公司 | Model joint training method and device based on block chain |
CN111062044A (en) * | 2019-12-09 | 2020-04-24 | 支付宝(杭州)信息技术有限公司 | Model joint training method and device based on block chain |
US11748835B2 (en) | 2020-01-27 | 2023-09-05 | Hewlett Packard Enterprise Development Lp | Systems and methods for monetizing data in decentralized model building for machine learning using a blockchain |
CN113767386B (en) * | 2020-01-27 | 2023-06-06 | 慧与发展有限责任合伙企业 | Network node operating in a distributed network |
US11876891B2 (en) | 2020-01-27 | 2024-01-16 | Hewlett Packard Enterprise Development Lp | Secure parameter merging using homomorphic encryption for swarm learning |
US11887204B2 (en) | 2020-01-27 | 2024-01-30 | Hewlett Packard Enterprise Development Lp | Systems and methods for monetizing data in decentralized model building for machine learning using a blockchain |
CN113767386A (en) * | 2020-01-27 | 2021-12-07 | 慧与发展有限责任合伙企业 | Security parameter merging using homomorphic encryption for group learning |
CN111859415A (en) * | 2020-06-18 | 2020-10-30 | 上海艾麒信息科技有限公司 | Neural network model encryption system and method |
CN111753984A (en) * | 2020-06-28 | 2020-10-09 | 中国银行股份有限公司 | Distributed AI training method, device and system based on block chain |
WO2021174779A1 (en) * | 2020-07-30 | 2021-09-10 | 平安科技(深圳)有限公司 | Data pre-processing system and method, computer device, and readable storage medium |
CN111882074A (en) * | 2020-07-30 | 2020-11-03 | 平安科技(深圳)有限公司 | Data preprocessing system, method, computer device and readable storage medium |
CN112699414A (en) * | 2020-08-14 | 2021-04-23 | 武汉乾云软件开发中心(有限合伙) | Block chain-based machine learning data evidence storage method and system |
CN112699414B (en) * | 2020-08-14 | 2024-03-15 | 武汉乾云软件开发中心(有限合伙) | Machine learning data evidence-storing method and system based on blockchain |
CN112016911A (en) * | 2020-08-20 | 2020-12-01 | 王红根 | Information management method based on block chain digital currency finance and digital service platform |
CN112163034A (en) * | 2020-09-03 | 2021-01-01 | 中国联合网络通信集团有限公司 | Model sharing method, node and block chain system |
CN112348063A (en) * | 2020-10-27 | 2021-02-09 | 广东电网有限责任公司电力调度控制中心 | Model training method and device based on federal transfer learning in Internet of things |
CN112801134A (en) * | 2020-12-30 | 2021-05-14 | 杭州趣链科技有限公司 | Gesture recognition model training and distributing method and device based on block chain and image |
CN113420017A (en) * | 2021-06-21 | 2021-09-21 | 上海特高信息技术有限公司 | Block chain application method for acquiring robot navigation algorithm training data set |
CN113420017B (en) * | 2021-06-21 | 2023-10-13 | 上海特高信息技术有限公司 | Block chain application method for acquiring training data set of robot navigation algorithm |
CN114037086B (en) * | 2021-10-19 | 2023-01-24 | 中国联合网络通信集团有限公司 | Block chain-based machine learning task distribution method, equipment and system |
CN114037086A (en) * | 2021-10-19 | 2022-02-11 | 中国联合网络通信集团有限公司 | Block chain-based machine learning task distribution method, equipment and system |
CN114707663A (en) * | 2022-04-15 | 2022-07-05 | 中国电信股份有限公司 | Distributed machine learning method and device, electronic equipment and storage medium |
CN114707663B (en) * | 2022-04-15 | 2024-02-23 | 中国电信股份有限公司 | Distributed machine learning method, device, electronic equipment and storage medium |
CN116566846B (en) * | 2023-07-05 | 2023-09-22 | 中国电信股份有限公司 | Model management method and system, shared node and network node |
CN116566846A (en) * | 2023-07-05 | 2023-08-08 | 中国电信股份有限公司 | Model management method and system, shared node and network node |
Also Published As
Publication number | Publication date |
---|---|
CN110490305B (en) | 2021-06-04 |
CN110991622B (en) | 2021-06-04 |
CN110991622A (en) | 2020-04-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110490305A (en) | Machine learning model processing method and node based on block chain network | |
US20220092586A1 (en) | Off network identity tracking in anonymous cryptocurrency exchange networks | |
CN110599181B (en) | Data processing method, device and equipment based on block chain and storage medium | |
US11689362B2 (en) | Distributed ledger for generating and verifying random sequence | |
CN110532323A (en) | Pupilage information processing method, device, electronic equipment and storage medium in block chain network | |
US20190287026A1 (en) | Learning service blockchain | |
US10554407B1 (en) | Protocols for decentralized networks | |
CN110599069B (en) | Application evaluation method and device based on block chain network | |
CN110599213B (en) | Article management method and device based on blockchain network and electronic equipment | |
Pasdar et al. | Connect api with blockchain: A survey on blockchain oracle implementation | |
CN110348830A (en) | A kind of network system based on block chain, authentication method, equipment and storage medium | |
CN110516474A (en) | User information processing method, device, electronic equipment and storage medium in block chain network | |
CN109450910A (en) | Data sharing method, data sharing network and electronic equipment based on block chain | |
CN110505239A (en) | A kind of information processing method based on block chain network, device and electronic equipment | |
US11902448B2 (en) | Hybrid blockchain architecture with computing pool | |
Pasdar et al. | Blockchain oracle design patterns | |
Shivers | Toward a secure and decentralized blockchain-based ride-hailing platform for autonomous vehicles | |
CN110517020A (en) | The node of project development method and block chain network based on block chain network | |
CN112163954B (en) | Transaction method and device based on block chain | |
US20220027803A1 (en) | Sustainable tokens for supply chain with privacy preserving protocol | |
CN109739827A (en) | A kind of block chain storage system based on double-strand framework | |
KR102051257B1 (en) | System for preventing counterfeit of non face-to-face transactions based on blockchain | |
Mustafa et al. | A governance framework with permissioned blockchain for the transparency in e-tendering process | |
CN116894727A (en) | Data processing method and device based on block chain and related equipment | |
Sun et al. | An efficient and secure trading framework for shared charging service based on multiple consortium blockchains |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40015683 Country of ref document: HK |
|
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |