CN109934628B - Feature processing method and device - Google Patents
Feature processing method and device Download PDFInfo
- Publication number
- CN109934628B CN109934628B CN201910176079.6A CN201910176079A CN109934628B CN 109934628 B CN109934628 B CN 109934628B CN 201910176079 A CN201910176079 A CN 201910176079A CN 109934628 B CN109934628 B CN 109934628B
- Authority
- CN
- China
- Prior art keywords
- feature
- type
- bits
- hash value
- processing method
- 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.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 15
- 239000013598 vector Substances 0.000 claims abstract description 27
- 238000000034 method Methods 0.000 claims abstract description 26
- 238000012545 processing Methods 0.000 claims abstract description 23
- 238000004364 calculation method Methods 0.000 claims description 12
- 230000000694 effects Effects 0.000 abstract description 3
- 230000009286 beneficial effect Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Images
Landscapes
- Storage Device Security (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present disclosure relates to a feature processing method and apparatus. According to one embodiment of the disclosure, the method comprises: assigning an object ID to the object; assigning a type ID to the type of the feature of the object; generating a hash value corresponding to the feature of the object using an operator corresponding to the type; and generating a signature representing a characteristic of the object based on the object ID, the type ID, and the hash value of the object. The method and the device of the disclosure have at least one of the following beneficial technical effects: the expression capability of the feature vector is enriched, and the feature processing process is ensured to have better usability and expandability.
Description
Technical Field
The present disclosure relates to computer technologies, and in particular, to a feature processing method and apparatus.
Background
Feature processing is an important part of the model training and service deployment process, and can be understood as a mapping process from attribute values (discrete attributes or continuous attributes) of users, contexts, articles, and the like to a number of feature vectors. A single feature is easy to represent, but in the process of model construction, a large number of attribute values are needed, the number of the attribute values is different, and the finally formed vector dimension can be in the order of ten thousand or even billion, so that certain difficulties exist in the aspect of feature management and feature processing. In the process of feature processing and service deployment, in order to strictly ensure the consistency of the data processing process, strictly consistent logic modules are required to perform this operation, for example, a common class library or the same set of codes are used, so when a feature is newly added, code intervention is often required, and usability and expandability are limited.
Disclosure of Invention
A brief summary of the disclosure is provided below in order to provide a basic understanding of some aspects of the disclosure. It should be understood that this summary is not an exhaustive overview of the disclosure. It is not intended to identify key or critical elements of the disclosure or to delineate the scope of the disclosure. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description that is discussed later.
According to a first aspect of the present disclosure, there is provided a feature processing method including:
assigning an object ID to the object;
assigning a type ID to the type of the feature of the object;
generating a hash value corresponding to the feature of the object using an operator corresponding to the type; and
a signature representing a characteristic of the object is generated based on the object ID, the type ID, and the hash value of the object.
According to a second aspect of the present disclosure, there is provided a feature processing apparatus including:
a first allocation unit configured to allocate an object ID to an object;
a second assigning unit configured to assign a type ID to a type of a feature of the object;
a calculation unit configured to generate a hash value corresponding to a feature of an object using an operator corresponding to a type; and
a generation unit configured to generate a signature representing a feature of the object based on the object ID, the type ID, and the hash value of the object.
According to a third aspect of the present disclosure, there is provided a storage medium having stored thereon a program that implements the feature processing method of the first aspect described above.
The technical scheme of the disclosure has at least one of the following technical effects: the expression capability of the feature vector is enriched, and the feature processing process is ensured to have better usability and expandability.
Drawings
The disclosure may be better understood by reference to the following description taken in conjunction with the accompanying drawings, which are incorporated in and form a part of this specification, along with the following detailed description. In the drawings:
FIG. 1 is a schematic flow diagram of a feature processing method according to one embodiment of the present disclosure; and
fig. 2 is a block diagram of a feature processing apparatus according to one embodiment of the present disclosure.
Detailed Description
Exemplary embodiments of the present disclosure will be described hereinafter with reference to the accompanying drawings. In the interest of clarity and conciseness, not all features of an actual embodiment are described in the specification. It will of course be appreciated that in the development of any such actual embodiment, numerous implementation-specific decisions may be made to achieve the developers' specific goals, such as compliance with system-related and business-related constraints, which will vary from one implementation to another.
Here, it should be further noted that, in order to avoid obscuring the present disclosure with unnecessary details, only the device structure closely related to the scheme according to the present disclosure is shown in the drawings, and other details not so related to the present disclosure are omitted.
It is to be understood that the disclosure is not limited to the described embodiments, as described below with reference to the drawings. In this context, embodiments may be combined with each other, features may be replaced or borrowed between different embodiments, one or more features may be omitted in one embodiment, where feasible.
In the process of training a model and deploying services, the operation of feature processing is inevitably carried out, and for features with different attribute values and different types, the vector dimension formed by the features can reach the level of ten-thousand even billions; especially when new features are added, code intervention is required, resulting in limited ease and scalability of feature processing.
According to one embodiment of the disclosure, different features are divided according to an object, the features in the object are processed in a unit of the object, and for each feature, a hash value corresponding to the feature is formed; and forming a signature corresponding to the feature based on the object ID, the type ID and the hash value, wherein the process ensures that the feature processing process has better usability and expandability.
In particular, fig. 1 is a flow diagram of a feature processing method 100 according to one embodiment of the present disclosure.
At step S101, an object ID is assigned to the object.
The objects can be user objects, advertisement objects, content objects and the like, and the features are divided into different objects according to different attributes of the features; for example, the feature having the attribute of the user is classified into a user object, the feature having the attribute of the advertisement is classified into an advertisement object, and the feature having the attribute of the content is classified into a content feature, wherein the feature having the attribute of the user includes features of types such as gender, age, height, and the like, the feature having the attribute of the advertisement includes features of types such as advertisement type, advertisement duration, advertisement source, and the like, and the feature having the attribute of the content includes features of types such as content type, content region, content time, and the like. The object ID is represented by block ID, for example, the block ID of the user object is 1, the block ID of the advertisement object is 2, and the block ID of the content object is 3.
At step S102, a type ID is assigned to the type of feature of the object.
One object has at least one type of feature, such as features belonging to an advertisement object including three types of features of gender, age, and height, and a type ID is assigned to each type of feature, and the type ID is represented by slot ID, for example, slot ID 1 for the gender feature, slot ID 2 for the age feature, and slot ID 3 for the height feature.
At step S103, generating a hash value corresponding to the feature of the object using an operator corresponding to the type;
at step S104, a signature of a feature of the object is generated based on the object ID, the type ID, and the hash value of the object.
In this embodiment, different features are divided according to an object, the object is used as a unit, feature processing is performed on the features in the object, and a hash value corresponding to each feature is formed for each feature; and forming a signature corresponding to the feature based on the object ID, the type ID and the hash value, wherein the process ensures that the feature processing process has better usability and expandability.
For example, generating a hash value corresponding to a feature of an object using an operator corresponding to a type includes:
carrying out one-dimensional sparsification on the value of the feature by using an operator corresponding to the type to obtain a feature vector; and carrying out Hash calculation on the characteristic vector to obtain a Hash value. In this embodiment, the one-dimensional sparsification and hash calculation process depends on configuration and an operator, the configuration can assign an operator corresponding to each type of feature, and input the value of the feature in the type and the corresponding parameter to the operator, and the hash value is obtained after the value of the feature is calculated by the operator.
The value of the feature is subjected to one-dimensional sparsification by using an operator corresponding to the type, for example, the one-hot one-dimensional sparsification mode is adopted to perform one-dimensional sparsification on the feature, so that a feature vector is obtained. one-hot one-dimensional sparsification is one-hot encoding, and for each feature, if m possible values exist in the feature, the one-hot one-dimensional sparsification is carried out, so that m binary features are obtained, the features are mutually exclusive, only one activation is carried out at a time, the data are sparse, and the function of feature expansion is achieved to a certain extent. In the present disclosure, when one-dimensional thinning is performed on a feature, other methods of thinning may be used, and the one-hot one-dimensional thinning is not limited to this.
The hash calculation of the feature vector corresponding to the value of the feature may be performed by, for example, using a murmurmur hash method to obtain a hash value corresponding to the feature, where the hash value is a 64-bit integer, and obtain a value in the feature vector corresponding to the feature, where the value is a floating point value, and the hash value and the value in the feature vector corresponding to the feature form a value pair. For example, for a feature (user a, gender male), the hash value-2366580462178760140 is obtained by computing the feature vector by using the murmur hash, and if the value in the feature vector corresponding to the feature is 1.0, the output value pair is (-2366580462178760140, 1.0).
Generating a signature of a feature of an object based on the object ID, the type ID, and the hash value of the object may include, for example:
continuously extracting a first preset number of bits of binary numbers as a first data section from the lowest bits of the binary representation of the object ID;
continuously extracting a second preset number of bits of binary numbers as a second data sector from the lowest bits of the binary representation of the type ID;
continuously extracting a binary number of a third preset number of bits from the lowest bit of the binary representation of the hash value as a third data section; and
and splicing the first data section, the second data section and the third data section in a predetermined sequence to be used as a signature. For example, the predetermined order may be to sequentially splice the first data segment, the second data segment, and the third data segment as the signature.
The sum of the first preset digit, the second preset digit and the third preset digit is equal to 64, and the third preset digit is greater than 32. For example, the first predetermined number of bits is 6, the second predetermined number of bits is 10, and the third predetermined number of bits is 48.
In an illustrative example, a signature corresponding to a feature includes 64 bits of numbers, the first 6 bits are an object identifier, the middle 10 bits are a type identifier, and the last 48 bits are a hash space of the signature, which can express about two hundred and eighty trillion hash spaces and can deal with a majority of valued hash spaces of type features. For example, for a feature (user a, gender male), the object ID is 1, the type ID is 1, the value "male" is hashed to the number-2366580462178760140 and the value 1.0, and the last 6-bit number of the binary representation of the object ID, the last 10-bit number of the binary representation of the type ID, and the last 48-bit number of the number-2366580462178760140 are spliced into the signature 288854468109568564 in sequence. Finally, the value pair (288854468109568564,1.0) can be output to express the characteristics of the sex male. For each sample, a plurality of characteristics are included, a series of signatures are obtained to express the sample after the scheme is adopted, and the signatures contain object information, type information and discrete value information of the corresponding characteristics, so that the length of the characteristic vector is expanded to be more than billions or even more than billions, and the expression capability of the characteristic vector is greatly enriched.
In the above example, since the last 6-bit number of the object ID and the last 10-bit number of the type ID are extracted, the range of the object ID is 0 to 63, and the range of the type ID is 0 to 1023.
In the present disclosure, the feature processing procedure may be mapped to a configuration file, which may facilitate deployment and iteration. The calculation aiming at each type of characteristics has universality, the reusability of codes is promoted, and the complexity of code intervention is reduced; meanwhile, the above method standardizes the process of feature processing.
In a second aspect of the present disclosure, a feature processing apparatus is also provided. Fig. 2 is a block diagram of a feature processing apparatus 200 in one embodiment provided in accordance with the present disclosure. As shown in fig. 2, the feature processing apparatus 200 includes: a first distribution unit 201, a second distribution unit 202, a calculation unit 203 and a generation unit 204.
A first assigning unit 201 for assigning an object ID to the object.
The objects can be user objects, advertisement objects, content objects and the like, and the features are divided into different objects according to different attributes of the features; for example, the feature having the attribute of the user is classified into a user object, the feature having the attribute of the advertisement is classified into an advertisement object, and the feature having the attribute of the content is classified into a content feature, wherein the feature having the attribute of the user includes features of types such as gender, age, height, and the like, the feature having the attribute of the advertisement includes features of types such as advertisement type, advertisement duration, advertisement source, and the like, and the feature having the attribute of the content includes features of types such as content type, content region, content time, and the like. The object ID is represented by block ID, for example, the block ID of the user object is 1, the block ID of the advertisement object is 2, and the block ID of the content object is 3.
A second assigning unit 202 for assigning a type ID to the type of the feature of the object.
One object has at least one type of feature, such as features belonging to an advertisement object including three types of features of gender, age, and height, and a type ID is assigned to each type of feature, and the type ID is represented by slot ID, for example, slot ID 1 for the gender feature, slot ID 2 for the age feature, and slot ID 3 for the height feature.
A calculating unit 203 for generating a hash value corresponding to the feature of the object using an operator corresponding to the type.
A generating unit 204 for generating a signature of a feature of the object based on the object ID, the type ID, and the hash value of the object.
In this embodiment, different features are divided according to an object, the object is used as a unit, feature processing is performed on the features in the object, and a hash value corresponding to each feature is formed for each feature; and forming a signature corresponding to the feature based on the object ID, the type ID and the hash value, wherein the process ensures that the feature processing process has better usability and expandability.
For example, the calculation unit 203 is configured to:
carrying out one-dimensional sparsification on the value of the feature by using an operator corresponding to the type to obtain a feature vector; and carrying out Hash calculation on the characteristic vector to obtain a Hash value.
In this embodiment, the one-dimensional sparsification and hash calculation process depends on configuration and an operator, the configuration can assign an operator corresponding to each type of feature, and input the value of the feature in the type and the corresponding parameter to the operator, and the hash value is obtained after the value of the feature is calculated by the operator.
For example, the generating unit 204 is configured to:
continuously extracting a first preset number of bits of binary numbers as a first data section from the lowest bits of the binary representation of the object ID;
continuously extracting a second preset number of bits of binary numbers as a second data sector from the lowest bits of the binary representation of the type ID;
continuously extracting a binary number of a third preset number of bits from the lowest bit of the binary representation of the hash value as a third data section; and
and splicing the first data section, the second data section and the third data section in a predetermined sequence to be used as a signature. For example, the predetermined order may be to sequentially splice the first data segment, the second data segment, and the third data segment as the signature.
The sum of the first preset digit, the second preset digit and the third preset digit is equal to 64, and the third preset digit is greater than 32. For example, the first predetermined number of bits is 6, the second predetermined number of bits is 10, and the third predetermined number of bits is 48.
In an illustrative example, a signature corresponding to a feature includes 64 bits of numbers, the first 6 bits are an object identifier, the middle 10 bits are a type identifier, and the last 48 bits are a hash space of the signature, which can express about two hundred and eighty trillion hash spaces and can deal with a majority of valued hash spaces of type features. For example, for a feature (user a, gender male), the object ID is 1, the type ID is 1, the value "male" is hashed to the number-2366580462178760140 and the value 1.0, and the last 6-bit number of the binary representation of the object ID, the last 10-bit number of the binary representation of the type ID, and the last 48-bit number of the number-2366580462178760140 are spliced into the signature 288854468109568564 in sequence. Finally, the value pair (288854468109568564,1.0) can be output to express the characteristics of the sex male.
In a third aspect of the present disclosure, there is provided a storage medium having stored thereon a program for implementing the feature processing method of the first aspect described above
By way of example, computer-readable storage media can be any available media that can be accessed by a computer or a data storage device, such as a server, data center, etc., that includes one or more available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others).
In the several embodiments provided in the present disclosure, it should be understood that the disclosed apparatus and method may be implemented in other manners. The above-described embodiments of the apparatus are merely illustrative, and for example, a module or a unit may be divided into only one logical function, and may be implemented in other ways, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted or not executed, where possible.
In addition, functional units in the embodiments of the present disclosure may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The method realizes one or more than one of the following technical effects:
1. the feature management function is to abstract the feature vector into a series of value pairs, each value pair includes a 64-bit integer (called a signature) and a floating-point value, such as a value pair (24343434323423423,1.0),24343434323423423 is a signature at a certain position in the feature vector, and 1.0 is a specific value at a certain position in the feature vector. Therefore, the length of the feature vector is expanded to be more than billions or even billions, and the expression capability of the feature vector is greatly enriched.
2. The plug-in implementation is that the feature extraction process is abstracted into an engine, the engine executes operators, and the operators and parameters are obtained by reading configuration, that is, the engine reads the configuration to define and execute corresponding extraction logic, so that what features are specifically extracted is only related to the configuration, and feature change and iteration are facilitated.
3. The concept of the block identifies the whole input, in the deep learning model, a plurality of feature vectors are often needed as input, the concept of the block (block) is introduced to identify the whole input, each block can contain a plurality of slots (slots), and the whole can be used as one feature vector to input the model. For example, the user's characteristic may be taken as a block in its entirety, with the gender characteristic being only one of the slots. Within the input of the model, we can have multiple blocks, such as user features, content features, advertisement features, etc., each of which can be expressed as a separate feature vector.
4. The slot (slot) concept is introduced, wherein a slot represents a characteristic slot, for example, gender is a slot, but the value of the gender corresponds to a specific signature (sign), the final signature is a 64-bit number and can be divided into three parts: the first 6 bits are the identification of the block (block), so the block ID needs to range from 0 to 63; the next 10 bits are the identification of the slot (slot), so the range of slot ID needs to be in the range of 0 to 1023; the final 48 bits are a signed hash space, which can express about two hundred and eighty trillion hash spaces, which is enough to deal with the valued hash space of most slot features, and the collision space is extremely small in probability.
Although a plurality of embodiments of the present disclosure have been described above, the scope of the present disclosure is not limited to these embodiments, and any person skilled in the art can easily conceive of changes or substitutions without departing from the present disclosure, and the corresponding aspects of the changes or substitutions should be covered by the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the claims.
Claims (8)
1. A method of feature processing, comprising:
assigning an object ID to the object;
assigning a type ID to the type of the feature of the object;
generating a hash value corresponding to a feature of the object using an operator corresponding to the type; and
generating a signature representing a feature of the object based on the object ID, the type ID, and the hash value of the object, comprising:
continuously extracting a first preset number of bits of binary numbers as a first data section from the lowest bits of the binary representation of the object ID;
continuously extracting a second preset number of bits of binary numbers as a second data sector from the lowest bits of the binary representation of the type ID;
continuously extracting a binary number of a third preset number of bits from the lowest bit of the binary representation of the hash value as a third data section; and
splicing the first data segment, the second data segment, and the third data segment in a predetermined order as the signature.
2. The feature processing method of claim 1, wherein generating a hash value corresponding to a feature of the object using an operator corresponding to the type comprises:
carrying out one-dimensional sparsification on the value of the feature by using an operator corresponding to the type to obtain a feature vector; and
and carrying out Hash calculation on the characteristic vector to obtain the Hash value.
3. The feature processing method according to claim 2, wherein the one-dimensional sparsification includes onehot one-dimensional sparsification.
4. The feature processing method according to claim 2, wherein the hash calculation includes a murmurr hash calculation.
5. The feature processing method according to claim 1, wherein the object ID has a value in a range of 0 to 63, and the type ID has a value in a range of 0 to 1023.
6. The feature processing method according to claim 1,
the sum of the first preset digit, the second preset digit and the third preset digit is equal to 64, and
the third preset number of bits is greater than 32.
7. A feature processing apparatus comprising:
a first allocation unit configured to allocate an object ID to an object;
a second assigning unit configured to assign a type ID to a type of a feature of the object;
a computing unit configured to generate a hash value corresponding to a feature of the object using an operator corresponding to the type; and
a generation unit configured to generate a signature representing a feature of the object based on the object ID, the type ID, and the hash value of the object;
the generation unit is further configured to:
continuously extracting a first preset number of bits of binary numbers as a first data section from the lowest bits of the binary representation of the object ID;
continuously extracting a second preset number of bits of binary numbers as a second data sector from the lowest bits of the binary representation of the type ID;
continuously extracting a binary number of a third preset number of bits from the lowest bit of the binary representation of the hash value as a third data section; and
splicing the first data segment, the second data segment, and the third data segment in a predetermined order as the signature.
8. A storage medium having stored thereon a program for implementing the feature processing method of any one of claims 1 to 5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910176079.6A CN109934628B (en) | 2019-03-08 | 2019-03-08 | Feature processing method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910176079.6A CN109934628B (en) | 2019-03-08 | 2019-03-08 | Feature processing method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109934628A CN109934628A (en) | 2019-06-25 |
CN109934628B true CN109934628B (en) | 2021-03-19 |
Family
ID=66986525
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910176079.6A Active CN109934628B (en) | 2019-03-08 | 2019-03-08 | Feature processing method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109934628B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103810551A (en) * | 2012-11-09 | 2014-05-21 | 金蝶软件(中国)有限公司 | Organizational unit object generation method and device in ERP system |
KR20160112862A (en) * | 2015-03-20 | 2016-09-28 | 이영희 | Slot machine game providing system based on user mobile device, and method thereof |
CN106407201A (en) * | 2015-07-29 | 2017-02-15 | 腾讯科技(深圳)有限公司 | Data processing method and apparatus |
CN107025233A (en) * | 2016-01-29 | 2017-08-08 | 苏宁云商集团股份有限公司 | A kind of processing method and processing device of data characteristics |
CN107291454A (en) * | 2017-05-27 | 2017-10-24 | 北京思特奇信息技术股份有限公司 | A kind of method and Commentary Systems that comment is added in the comment list of event |
CN109409030A (en) * | 2018-10-19 | 2019-03-01 | 微梦创科网络科技(中国)有限公司 | A kind of digital publishing rights generation method and device |
US10540338B2 (en) * | 2017-01-30 | 2020-01-21 | Alfresco Software, Inc. | Scalable fine grained access control within a search engine |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU2015258203A1 (en) * | 2009-06-26 | 2015-12-03 | Hbgary, Inc. | Fuzzy hash algorithm |
CN108108436B (en) * | 2017-12-20 | 2020-07-31 | 东软集团股份有限公司 | Data storage method and device, storage medium and electronic equipment |
CN109146083B (en) * | 2018-08-06 | 2021-07-23 | 创新先进技术有限公司 | Feature encoding method and apparatus |
-
2019
- 2019-03-08 CN CN201910176079.6A patent/CN109934628B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103810551A (en) * | 2012-11-09 | 2014-05-21 | 金蝶软件(中国)有限公司 | Organizational unit object generation method and device in ERP system |
KR20160112862A (en) * | 2015-03-20 | 2016-09-28 | 이영희 | Slot machine game providing system based on user mobile device, and method thereof |
CN106407201A (en) * | 2015-07-29 | 2017-02-15 | 腾讯科技(深圳)有限公司 | Data processing method and apparatus |
CN107025233A (en) * | 2016-01-29 | 2017-08-08 | 苏宁云商集团股份有限公司 | A kind of processing method and processing device of data characteristics |
US10540338B2 (en) * | 2017-01-30 | 2020-01-21 | Alfresco Software, Inc. | Scalable fine grained access control within a search engine |
CN107291454A (en) * | 2017-05-27 | 2017-10-24 | 北京思特奇信息技术股份有限公司 | A kind of method and Commentary Systems that comment is added in the comment list of event |
CN109409030A (en) * | 2018-10-19 | 2019-03-01 | 微梦创科网络科技(中国)有限公司 | A kind of digital publishing rights generation method and device |
Non-Patent Citations (1)
Title |
---|
FDSS:基于HDFS的海量音乐特征数据存储***;范远超 等;《计算机研究与发展》;20120626;正文第2节 * |
Also Published As
Publication number | Publication date |
---|---|
CN109934628A (en) | 2019-06-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10778441B2 (en) | Redactable document signatures | |
CN104008064B (en) | The method and system compressed for multi-level store | |
CN110825363B (en) | Intelligent contract acquisition method and device, electronic equipment and storage medium | |
CN107391549B (en) | Artificial intelligence based news recall method, device, equipment and storage medium | |
CN111177438B (en) | Image characteristic value searching method and device, electronic equipment and storage medium | |
US9619657B2 (en) | Method and apparatus for storing redeem code, and method and apparatus for verifying redeem code | |
KR101930293B1 (en) | Apparatus and Method for Identifying Variety Malicious Code Using Static Analysis and Dynamic Analysis | |
US20180039824A1 (en) | Clustering large database of images using multilevel clustering approach for optimized face recognition process | |
JP2015501017A (en) | Image search method | |
CN109978006B (en) | Face image clustering method and device | |
KR20140121832A (en) | Filtering redundant consumer transaction rules | |
CN109934628B (en) | Feature processing method and device | |
CN113742332A (en) | Data storage method, device, equipment and storage medium | |
EP3232319B1 (en) | Random number generation apparatus, random number generation method and program | |
KR102225820B1 (en) | Apparatus and method for generating hybrid feature vector for machine learning for malicious code analysis | |
KR102197993B1 (en) | Data distribution storage apparatus and method using partitioned iteration code | |
JP6123372B2 (en) | Information processing system, name identification method and program | |
CN113392138A (en) | Statistical analysis method and device for private data, server and storage medium | |
CN112632054A (en) | Data set duplication removing method based on attribute encryption, storage medium and system | |
CN111176830A (en) | Information flow distribution method and device and server system | |
CN113641964B (en) | Repackaging application detection method, electronic device and storage medium | |
KR102289411B1 (en) | Weighted feature vector generation device and method | |
JP7497734B2 (en) | Graph search device, graph search method, and program | |
CN116204201B (en) | Service processing method and device | |
US11521409B2 (en) | Hybrid clustering and pairwise comparison document matching |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |