CN113656696A - Session recommendation method and device - Google Patents
Session recommendation method and device Download PDFInfo
- Publication number
- CN113656696A CN113656696A CN202110973346.XA CN202110973346A CN113656696A CN 113656696 A CN113656696 A CN 113656696A CN 202110973346 A CN202110973346 A CN 202110973346A CN 113656696 A CN113656696 A CN 113656696A
- Authority
- CN
- China
- Prior art keywords
- session
- vector
- node
- recommendation
- graph
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 66
- 239000013598 vector Substances 0.000 claims abstract description 172
- 230000006870 function Effects 0.000 claims description 30
- 238000013528 artificial neural network Methods 0.000 claims description 19
- 230000015654 memory Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 11
- 238000012856 packing Methods 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 8
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000007547 defect Effects 0.000 abstract description 5
- 210000005036 nerve Anatomy 0.000 abstract description 5
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 230000000875 corresponding effect Effects 0.000 description 12
- 230000007246 mechanism Effects 0.000 description 10
- 230000008569 process Effects 0.000 description 8
- 238000004891 communication Methods 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 7
- 238000009826 distribution Methods 0.000 description 7
- 238000012545 processing Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000002457 bidirectional effect Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000007704 transition Effects 0.000 description 3
- 208000003443 Unconsciousness Diseases 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000000750 progressive effect Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 241000282412 Homo Species 0.000 description 1
- 108010001267 Protein Subunits Proteins 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007177 brain activity Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 230000000739 chaotic effect Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 230000003920 cognitive function Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000001339 gustatory effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000000638 stimulation Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Probability & Statistics with Applications (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to the technical field of artificial intelligence, and provides a session recommendation method and a device, wherein the session recommendation method comprises the following steps: converting a plurality of conversation sequences into a conversation graph; generating a vector representation of the current node according to a forward node hidden vector and a backward node hidden vector of the current node in each session graph; and selecting a recommendation session from the plurality of session sequences according to the vector representation, and recommending to the user. The invention overcomes the defects that the direction information of non-adjacent complex conversations cannot be learned and the recommendation performance is reduced due to the introduction of information redundancy in the prior art, and provides a conversation recommendation method based on a digraph nerve. The problem of being unable to identify useful direction information and useful connections between nodes in non-adjacent session networks is solved. The best performance is achieved in the disclosed session recommendation task, while the learned session representation is very robust to session length and shows the best performance in both long and short sessions.
Description
Technical Field
The application belongs to the technical field of artificial intelligence, and particularly relates to a session recommendation method and device.
Background
In the prior art of conversational recommendation, the conventional conversational recommendation method generally predicts the next click item by mapping the current conversation to a markov chain. This approach is not suitable for anonymous session recommendations because the markov-based approach only considers the order transition of two adjacent terms. The recurrent neural network model uses nonlinear transitions between successive hidden states to model causal factors in a conversational sequence. Multi-layer gated cyclic units, such as GRU4REC, exploit their complex hidden dynamics to capture long-term dependencies. NARM utilizes an attention mechanism to integrate global and local encoders, improving from an item conversion perspective. The STAMP employs an attention-based short-term memory network to capture the current interest of the session.
The markov chain attribute is translated to assume that the next user operation depends only on a limited number of recent operations, and the markov attribute limits the historical dependency of the conversational process. Markov-based and recurrent neural network-based approaches only model the predictive modeling of event transitions for adjacent time series of conversations, but do not enable conversational representation learning for non-adjacent complex conversations. The existing neural methods of the graph have a common limitation that only homographs with fixed directions are processed in the conversation recommendation, and the potential correlation between the diversity of node or edge types and the direction features is ignored. The construction of the graph is not simple, especially when a session has multiple nodes and edge types representing complex relationships. The existing method ignores the existence of node or edge type difference and uniformly expresses the graph. This representation is not optimal due to the lack of type information. These characteristics can significantly affect the accuracy of the session-based recommendation task.
Disclosure of Invention
The invention can be used in the technical field of applying artificial intelligence technology in the aspect of finance, and can also be used in any field except the financial field, and the technical defect that useful direction information and useful connection between nodes cannot be identified in a non-adjacent session network is overcome. The best performance is achieved in the disclosed session recommendation task, while the learned session representation is very robust to session length and shows the best performance in both long and short sessions.
In order to solve the technical problems, the invention provides the following technical scheme:
in a first aspect, the present invention provides a session recommendation method, including:
converting a plurality of conversation sequences into a conversation graph;
generating a vector representation of the current node according to a forward node hidden vector and a backward node hidden vector of the current node in each session graph;
and selecting a recommendation session from the plurality of session sequences according to the vector representation, and recommending to the user.
In one embodiment, the generating a vector representation of the current node according to the forward node hidden vector and the backward node hidden vector of the current node in each session graph includes:
the forward node hidden vectors and the backward node hidden vectors are learned bi-directionally using a bi-directional graph neural network to generate the vector representation.
In one embodiment, the bi-directionally learning the forward node hidden vectors and the backward node hidden vectors using a bi-directional neural network to generate the vector representation includes:
generating an updating function according to the node vector, the weight, the bias parameter and the corresponding out-degree/in-degree matrix of the session graph;
learning the forward node hidden vector and the backward node hidden vector according to the update function to generate the vector representation.
In one embodiment, the selecting a recommendation session from the plurality of session sequences according to the vector representation and recommending to the user includes:
packing the vector representation of all the nodes of each meeting picture to obtain the vector representation of the session corresponding to the current session picture;
scoring the vector representation for each session according to the user's session history and current preferences to select the recommended sessions from the plurality of sequences of sessions for recommendation to the user.
In a second aspect, the present invention provides a conversation recommendation apparatus, including:
the conversation sequence conversion module is used for converting a plurality of conversation sequences into a conversation graph;
a node vector representation generating module, configured to generate a vector representation of a current node according to a forward node hidden vector and a backward node hidden vector of the current node in each session graph;
and the recommendation session selection module is used for selecting a recommendation session from the plurality of session sequences according to the vector representation and recommending the recommendation session to a user.
In one embodiment, the node vector representation generating module includes:
a node vector generation unit configured to bidirectionally learn the forward node hidden vector and the backward node hidden vector using a bi-directional neural network to generate the vector representation.
In one embodiment, the node vector generating unit includes:
the updating function generating unit is used for generating an updating function according to the node vector, the weight, the bias parameter and the corresponding out-degree/in-degree matrix of the session graph;
a node vector representation generating unit, configured to learn the forward node hidden vector and the backward node hidden vector according to the update function to generate the vector representation.
In one embodiment, the recommendation session selection module includes:
the vector representation packing unit is used for packing the vector representations of all the nodes of each meeting picture to obtain the vector representation of the session corresponding to the current session graph;
and the recommendation session selection unit is used for grading the vector representation of each session according to the session history of the user and the current preference so as to select the recommendation session from the plurality of session sequences and recommend the recommendation session to the user.
In a third aspect, the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the session recommendation method when executing the program.
In a fourth aspect, the invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of a session recommendation method.
As can be seen from the above description, an embodiment of the present invention provides a method and an apparatus for session recommendation, which first convert a plurality of session sequences into a session graph; then, learning a forward node hidden vector and a backward node hidden vector of the current node in each session graph to generate a vector representation of the current node; and finally, selecting a recommendation session from the plurality of session sequences according to the vector representation and recommending to the user. The invention overcomes the defects that the direction information of non-adjacent complex conversations cannot be learned and the recommendation performance is reduced due to the introduction of information redundancy in the prior art, and provides a conversation recommendation method based on a digraph nerve. The problem of being unable to identify useful direction information and useful connections between nodes in non-adjacent session networks is solved. The best performance is achieved in the disclosed session recommendation task, while the learned session representation is very robust to session length and shows the best performance in both long and short sessions.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flowchart illustrating a session recommendation method according to an embodiment of the present invention;
FIG. 2 is a flow chart illustrating step 200 according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating step 201 according to an embodiment of the present invention;
FIG. 4 is a flowchart illustrating step 300 according to an embodiment of the present invention;
FIG. 5 is a flowchart illustrating a session recommendation method according to an embodiment of the present invention;
FIG. 6 is a flowchart illustrating node vector learning according to an embodiment of the present invention;
FIG. 7 is a block diagram of a session recommendation device in an embodiment of the present invention;
FIG. 8 is a block diagram of the node vector representation generation module 20 according to an embodiment of the present invention;
FIG. 9 is a block diagram of a node vector generation unit 201 according to an embodiment of the present invention;
FIG. 10 is a block diagram of the recommendation session selection module 30 in an embodiment of the present invention;
fig. 11 is a schematic structural diagram of an electronic device in an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
It should be noted that the terms "comprises" and "comprising," and any variations thereof, in the description and claims of this application and the above-described drawings, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.
The embodiment of the present invention provides a specific implementation manner of a session recommendation method, and referring to fig. 1, the method specifically includes the following contents:
step 100: a plurality of conversation sequences is converted into a conversation graph.
In particular, all conversation sequences are modeled as a conversation graph. A conversation sequence refers to a sequential collection of items used in one interactive transaction of a user.
Step 200: and generating a vector representation of the current node according to the forward node hidden vector and the backward node hidden vector of the current node in each session graph.
Specifically, for each session graph, a bi-directional graph neural network is utilized to pass through the GNNs separatelyRAnd GNNLAnd learning a forward node hidden vector and a backward node hidden vector.
Step 300: and selecting a recommendation session from the plurality of session sequences according to the vector representation, and recommending to the user.
Although the conventional deep learning method has been applied with great success in extracting features of euclidean space data, data in many practical application scenarios are generated from non-euclidean space, and the performance of the conventional deep learning method in processing the non-euclidean space data is still difficult to satisfy. For example, in e-commerce, a Graph (Graph) based learning system can make very accurate recommendations using the interaction between users and products, but the complexity of graphs makes existing deep learning algorithms face significant challenges in processing. This is because graphs are irregular, each with a variable sized chaotic node, and each node in the graph has a different number of neighboring nodes, resulting in some important operations (e.g., convolution) that are easily computed on the Image (Image) but are no longer suitable for direct graph use. Furthermore, one core assumption of existing deep learning algorithms is that data samples are independent of each other. However, this is not the case for graphs, where each data sample (node) in the graph has edges that are related to other real data samples (nodes) in the graph, and this information can be used to capture the interdependencies between instances. In recent years, there has been an increasing interest in the expansion of deep learning methods on graphs. Under the drive of the success of multiple factors, researchers have defined and designed a Neural network structure for processing Graph data by taking the ideas of a convolutional network, a cyclic network and a depth automatic encoder as reference, so that a new research hotspot is 'Graph Neural Networks (GNN)'.
On the other hand, conversational recommendation is widely used in recommendation tasks due to its outstanding advantages in privacy protection. Given the limitations of user privacy data protection, session recommendations may improve the user experience or enable accurate marketing without using user data. The session recommendation is to recommend the next possible click event based on the historical session sequence. The input of the conversational recommendation is a conversational sequence and the output is the probability of all possible click events, wherein the probability value comprises the score of the potential recommended item.
As can be seen from the above description, an embodiment of the present invention provides a session recommendation method, which first converts a plurality of session sequences into a session graph; then, learning a forward node hidden vector and a backward node hidden vector of the current node in each session graph to generate a vector representation of the current node; and finally, selecting a recommendation session from the plurality of session sequences according to the vector representation and recommending to the user. The invention overcomes the defects that the direction information of non-adjacent complex conversations cannot be learned and the recommendation performance is reduced due to the introduction of information redundancy in the prior art, and provides a conversation recommendation method based on a digraph nerve. The problem of being unable to identify useful direction information and useful connections between nodes in non-adjacent session networks is solved. The best performance is achieved in the disclosed session recommendation task, while the learned session representation is very robust to session length and shows the best performance in both long and short sessions.
In one embodiment, referring to fig. 2, step 200 further comprises:
step 201: the forward node hidden vectors and the backward node hidden vectors are learned bi-directionally using a bi-directional graph neural network to generate the vector representation.
It is understood that Graph Neural Networks (GNNs) are a class of deep learning-based methods of processing graph domain information. The figures mentioned herein all refer to the figure in Graph theory (Graph). It is a graph composed of several nodes and edges connecting two nodes, and is used to depict the relationship between different nodes. Given a graph G, each node has its own feature (feature), with xvA feature representing node v; the edge connecting two nodes also has its own characteristic, with x(v,u)A feature representing an edge between node v and node u; the goal of the GNN learning is to obtain the graph-perceived hidden state h of each nodev(state embedding), for each node, its hidden state contains information from neighboring nodes. The GNN obtains an embedded representation of the current node by iteratively updating the hidden states of all nodes.
In one embodiment, referring to fig. 3, step 201 further includes:
step 2011: generating an updating function according to the node vector, the weight, the bias parameter and the corresponding out-degree/in-degree matrix of the session graph;
step 2012: learning the forward node hidden vector and the backward node hidden vector according to the update function to generate the vector representation.
The learning process of the node vector in the session graph specifically includes: in the figure GsTo obtain phiiNode v in the directions,iExpressed, the update function employed is as follows:
whereinAnd H and b are respectively weight and bias parameters. A. thesIs a matrix from D x 2DThe two columns to which the selected node v corresponds are the corresponding two columns, whereinIs the corresponding out/in matrix of the bipartite graph of each session, asIs a 1 x D-dimensional vector that represents the result of the interaction of the current node and neighboring nodes through edges.
Equations (2) - (5) are similar to the calculation process of GRU.Is in the direction phiiTo the final updated node state.
In one embodiment, referring to fig. 4, step 300 further comprises:
step 301: packing the vector representation of all the nodes of each meeting picture to obtain the vector representation of the session corresponding to the current session picture;
step 302: scoring the vector representation for each session according to the user's session history and current preferences to select the recommended sessions from the plurality of sequences of sessions for recommendation to the user.
In another embodiment, before step 301, the forward node hidden vector and the backward node hidden vector may be summed by using an attention mechanism, and the node vectors are packed according to the session to obtain the session representation. Finally, the likelihood of each item is predicted by combining the history of the session and the current preferences.
The attention mechanism, also known as attention mechanism, may enable a neural network to have the ability to focus on a subset of its inputs (or features): a particular input is selected. Attention may be applied to any type of input regardless of its shape. In the case of limited computing power, an attention mechanism (attention mechanism) is a resource allocation scheme of a main means for solving the information overload problem, and computing resources are allocated to more important tasks.
Attention is generally divided into two categories: one is conscious attention from top to bottom, called focused (focus) attention. Focused attention refers to attention that has a predetermined purpose, is task dependent, and is actively focused on a subject consciously; the other is a bottom-up unconscious attention called salience-based attention. Attention based on significance is attention driven by external stimuli, does not require active intervention, and is also task independent. If a subject's stimulation information differs from its surrounding information, an unconscious "winner-take-all" or gating (gating) mechanism may divert attention to the subject. Regardless of whether such attention is intended or unintended, most human brain activities require attention, such as memorizing information, reading or thinking, and the like.
In cognitive neurology, attention is an indispensable complex cognitive function of humans, meaning the ability of a person to select to ignore some information while focusing on others. In daily life, a large amount of sensory input is received through visual, auditory, tactile, and the like. The human brain can work in order in these outside information bombings because the human brain can intentionally or unintentionally select a small portion of useful information from these large amounts of input information to focus on and ignore other information. This ability is called attention. Attention may be expressed as external stimuli (auditory, visual, gustatory, etc.) or internal consciousness (thinking, recall, etc.).
Multi-head attention (multi-head attentions) uses multiple queries to compute multiple selections of information from input information in parallel. Each focusing on a different part of the input information. Hard attention, i.e. the expectation of all input information based on the attention distribution. There is also an attention to one location only, called hard attention (hardattention).
Hard attention has two implementations, one is to select the highest probability of input information. Another type of hard attention may be achieved by means of random sampling over the attention distribution. One drawback of hard attention is that the information is selected based on the way of maximum sampling or random sampling. The functional relationship between the final loss function and the attention distribution is therefore not derivable and therefore cannot be trained using a back propagation algorithm. To use back-propagation algorithms, soft attention is typically used instead of hard attention.
Key-value pair attention: more generally, input information may be represented in a key-value pair (key-value pair) format, where "keys" are used to calculate attention distributions and "values" are used to generate selected information.
Structural attention: to select task-related information from the input information, active attention is paid to a multinomial distribution over all the input information, which is a flat (flat) structure. If the input information itself has a hierarchical structure, such as text can be divided into different granularity levels of words, sentences, paragraphs, chapters, etc., the hierarchical attention can be used for better information selection. In addition, a graph model can be used to construct more complex structured attention distributions, assuming a context-dependent binomial distribution of attention.
As can be seen from the above description, an embodiment of the present invention provides a session recommendation method, which first converts a plurality of session sequences into a session graph; then, learning a forward node hidden vector and a backward node hidden vector of the current node in each session graph to generate a vector representation of the current node; and finally, selecting a recommendation session from the plurality of session sequences according to the vector representation and recommending to the user. Specifically, the invention has the following beneficial effects:
1. the invention provides a bipartite neural network for learning a session representation on a directed graph. The method converts an abnormal graph on a conversation into a bidirectional connection graph with different edge types, and simultaneously aggregates the learned directed graphs to represent the conversation. On a typical session recommendation data set, the bipartite graph structure learned by the method can obtain more effective session representation, and the accuracy (Precision), Recall (Recall) and average to number-rank-name (MRR) indexes of the bipartite graph structure exceed the optimal models in recent years. The method opens up a new way for the GNN optimization graph structure, so that the GNN can perform bidirectional learning operation on the session recommendation task.
2. Most GNN-based recommendations represent sessions as homogeneous compositions, ignoring the potential correlation between the diversity of node or edge types and directional features. The invention provides a digraph neural network model, which relates to identifying useful connection between nodes on a directed graph, and for each session graph, BiGNN can respectively capture hidden vectors of forward and backward nodes, and simultaneously learn soft selection of edge types and compound relations to obtain effective session representation. At the same time, the learned session representation is very robust to session length and shows the best performance in both long and short sessions.
3. The invention only needs the sequence information of the matters without knowing the data information of the user, has the outstanding advantages of privacy protection widely applied to the current recommendation task and becomes a part of the daily online user experience. In view of the limitation of user privacy data protection, the session recommendation can improve the user experience or realize accurate marketing without using user data, and the method achieves the most advanced achievements in the aspects of e-commerce, social networks and media streams.
In a specific embodiment, the present invention further provides a specific embodiment of a session recommendation method, which is shown in fig. 5 and specifically includes the following contents.
Description of terms:
and (3) architecture: in the graph neural network, the information of the structure and the nodes of the graph is used as output, and the output of the graph can be divided into the following types according to different graph analysis tasks:
and (3) node level output: the class output is correlated to the regression and classification of the nodes. Since graph convolution networks will give a potential representation of graph data nodes, typically a perceptual layer or softmax layer is added behind the GCN.
And (3) side-level output: the class output is related to the classification and connection prediction tasks of the edges. In order to be able to predict the connection strength of an edge, a function is additionally added, taking potential representations of two nodes as input.
And (3) output of a graph stage: this type of output is typically associated with the classification task of the graph. To obtain a more compact representation from a graph, a pooling layer may be used to compact a graph to generate a sub-graph (some nodes and edges may be removed).
Graph convolution:
(1) each node transmits the characteristic information of the node to the neighbor nodes
(2) Each node collects the characteristic information of the neighbor nodes and the self-characteristic information and fuses the local structures
(3) Similar to the activation function in the traditional deep learning, the activation function is added in the graph convolution, the nonlinear transformation is carried out on the information of the nodes, and the expression capability of the model is enhanced
(4) The key of the graph convolution network is to learn a function and collect the characteristic information of the current node and the characteristic information of the neighbor nodes.
S1: and performing information aggregation on the session graph.
First, a learning process of a node vector in a session graph is given. Referring to FIG. 6, in FIG. GsTo obtain phiiNode v in the directions,iExpressed, the update function employed is as follows:
whereinAnd H and b are respectively weight and bias parameters. A. thesIs a matrix from D x 2DThe two columns to which the selected node v corresponds are the corresponding two columns, whereinIs the corresponding out/in matrix of the bipartite graph of each session, asIs a 1 x D-dimensional vector that represents the result of the interaction of the current node and neighboring nodes through edges.
Equations (2) - (5) are similar to the calculation process of GRU.Is in the direction phiiTo the final updated node state.
S2: by GNNRAnd GNNLAnd performing bidirectional learning on the forward node hidden vector and the backward node hidden vector.
The calculation results in different directions under the two edge types are respectively considered. Consideration in computingThe in-degree and out-degree of the two-way information transmission are calculated and considered. Given set of directions { Φ1,…,ΦpGet P group specific node embeddingWith P groups of session-specific node embeddings represented at the node level as inputs, weights per directionAs follows:
wherein attsqA deep neural network for performing an attention mechanism on a directional layer is represented that can capture various types of directional information behind a conversational diagram. Node embedding and session level attention is then converted by metricsThe similarity between the vectors q determines the importance of the session-level embedding, and the average of all session-level node embedding importance is taken as the importance of each direction interpretation. The importance of each direction is defined as a, as follows:
wherein WqIs the weight matrix, b is the bias vector, and q is the session level attention vector. After the importance of each direction is obtained, the softmax function pair is usedAnd (6) carrying out normalization.
S3: a session representation.
For a session-based recommendation task, each direction may have a different weight. Using the learned weight as the coefficient of embedding specific to the conversation to obtain the final embedding viThe following are:
to better predict the scores of potential recommended items, long-term preferences and current interests are incorporated as a session embedding. For a session s ═ vs,1,vs,2,…,vs,|s|]Taking the last term v of a clicks,|s|As a locally embedded representation sl=vs,|s|. Then, global embedding s of the session graph is considered by collecting all node vectorsg. Considering that the embedded information may have different priorities, a soft attention mechanism is employed to better express global session preferences:
γi=WTσ(W1vs,|s|+W2vi+c) (10)
ss=W3[sl;sg] (12)
s4: and predicting conversation.
By expressing s of the conversationsEmbedding vector v with candidateiMultiplying, calculating a score for each candidate
wherein,a recommendation score representing all of the candidate items,representing the probability of the next click of the node in session s. For each session graph, a cross entropy loss function is established that minimizes true and predicted values, as follows:
as can be seen from the above description, an embodiment of the present invention provides a session recommendation method, which first converts a plurality of session sequences into a session graph; then, learning a forward node hidden vector and a backward node hidden vector of the current node in each session graph to generate a vector representation of the current node; and finally, selecting a recommendation session from the plurality of session sequences according to the vector representation and recommending to the user. The invention provides a session recommendation method based on digraph nerves, which comprises the following steps: graph session representations are learned in an end-to-end manner using a bipartite graph neural network (BiGNN). The BiGNN converts the session into a bipartite graph containing node and edge type information. As shown in fig. 1, we first model all conversation sequences as a conversation graph. For each session graph, the bipartite graph neural network learns forward and backward node hidden vectors by GNNR and GNNL, respectively. Then, an attention mechanism is adopted to carry out weighted summation on the forward hidden vector and the backward hidden vector, and the node vector is packed according to the conversation to obtain conversation representation. Finally, we predict the likelihood of each item by combining the history of the session and the current preferences.
Based on the same inventive concept, the embodiment of the present application further provides a session recommendation device, which can be used to implement the methods described in the foregoing embodiments, such as the following embodiments. Because the principle of the session recommendation device for solving the problem is similar to the session recommendation method, the implementation of the session recommendation device can be implemented by referring to the session recommendation method, and repeated details are not repeated. As used hereinafter, the term "unit" or "module" may be a combination of software and/or hardware that implements a predetermined function. While the system described in the embodiments below is preferably implemented in software, implementations in hardware, or a combination of software and hardware are also possible and contemplated.
An embodiment of the present invention provides a specific implementation manner of a session recommendation device capable of implementing a session recommendation method, and referring to fig. 7, the session recommendation device specifically includes the following contents:
a conversation sequence conversion module 10, configured to convert a plurality of conversation sequences into a conversation graph;
a node vector representation generating module 20, configured to generate a vector representation of a current node according to a forward node hidden vector and a backward node hidden vector of the current node in each session graph;
and the recommendation session selection module 30 is configured to select a recommendation session from the multiple session sequences according to the vector representation, and recommend the selected recommendation session to the user.
In one embodiment, referring to fig. 8, the node vector representation generating module 20 includes:
a node vector generating unit 201, configured to bidirectionally learn the forward node hidden vector and the backward node hidden vector by using a bidirectional graph neural network, so as to generate the vector representation.
In one embodiment, referring to fig. 9, the node vector generating unit 201 includes:
an update function generation unit 2011, configured to generate an update function according to the node vector, the weight, the bias parameter, and the corresponding out-degree/in-degree matrix of the session graph;
a node vector representation generating unit 2012 for learning the forward node hidden vector and the backward node hidden vector according to the update function to generate the vector representation.
In one embodiment, referring to fig. 10, the recommendation session selection module 30 includes:
a vector representation packing unit 301, configured to pack vector representations of all nodes of each meeting graph, to obtain a vector representation of a session corresponding to the current session graph;
a recommendation session selecting unit 302, configured to score the vector representation of each session according to the session history of the user and the current preference, so as to select the recommendation session from the multiple session sequences and recommend the recommendation session to the user.
As can be seen from the above description, an embodiment of the present invention provides a session recommendation apparatus, which first converts a plurality of session sequences into a session graph; then, learning a forward node hidden vector and a backward node hidden vector of the current node in each session graph to generate a vector representation of the current node; and finally, selecting a recommendation session from the plurality of session sequences according to the vector representation and recommending to the user. The invention overcomes the defects that the direction information of non-adjacent complex conversations cannot be learned and the recommendation performance is reduced due to the introduction of information redundancy in the prior art, and provides a conversation recommendation method based on a digraph nerve. The problem of being unable to identify useful direction information and useful connections between nodes in non-adjacent session networks is solved. The best performance is achieved in the disclosed session recommendation task, while the learned session representation is very robust to session length and shows the best performance in both long and short sessions.
An embodiment of the present application further provides a specific implementation manner of an electronic device, which is capable of implementing all steps in the session recommendation method in the foregoing embodiment, and referring to fig. 11, the electronic device specifically includes the following contents:
a processor (processor)1201, a memory (memory)1202, a communication Interface 1203, and a bus 1204;
the processor 1201, the memory 1202 and the communication interface 1203 complete communication with each other through the bus 1204; the communication interface 1203 is used for implementing information transmission between related devices such as server-side devices and client-side devices;
the processor 1201 is configured to call the computer program in the memory 1202, and the processor executes the computer program to implement all the steps in the session recommendation method in the above embodiments, for example, the processor executes the computer program to implement the following steps:
step 100: converting a plurality of conversation sequences into a conversation graph;
step 200: generating a vector representation of the current node according to a forward node hidden vector and a backward node hidden vector of the current node in each session graph;
step 300: and selecting a recommendation session from the plurality of session sequences according to the vector representation, and recommending to the user.
Embodiments of the present application further provide a computer-readable storage medium capable of implementing all steps in the session recommendation method in the above embodiments, where the computer-readable storage medium stores thereon a computer program, and the computer program, when executed by a processor, implements all steps of the session recommendation method in the above embodiments, for example, when the processor executes the computer program, the processor implements the following steps:
step 100: converting a plurality of conversation sequences into a conversation graph;
step 200: generating a vector representation of the current node according to a forward node hidden vector and a backward node hidden vector of the current node in each session graph;
step 300: and selecting a recommendation session from the plurality of session sequences according to the vector representation, and recommending to the user.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the hardware + program class embodiment, since it is substantially similar to the method embodiment, the description is simple, and the relevant points can be referred to the partial description of the method embodiment.
The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
Although the present application provides method steps as in an embodiment or a flowchart, more or fewer steps may be included based on conventional or non-inventive labor. The order of steps recited in the embodiments is merely one manner of performing the steps in a multitude of orders and does not represent the only order of execution. When an actual apparatus or client product executes, it may execute sequentially or in parallel (e.g., in the context of parallel processors or multi-threaded processing) according to the embodiments or methods shown in the figures.
For convenience of description, the above devices are described as being divided into various modules by functions, and are described separately. Of course, in implementing the embodiments of the present description, the functions of each module may be implemented in one or more software and/or hardware, or a module implementing the same function may be implemented by a combination of multiple sub-modules or sub-units, and the like. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may therefore be considered as a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.
In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.
The embodiments of this specification may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The described embodiments may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment. In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of an embodiment of the specification. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
The above description is only an example of the embodiments of the present disclosure, and is not intended to limit the embodiments of the present disclosure. Various modifications and variations to the embodiments described herein will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present specification should be included in the scope of the claims of the embodiments of the present specification.
Claims (10)
1. A method for session recommendation, comprising:
converting a plurality of conversation sequences into a conversation graph;
generating a vector representation of the current node according to a forward node hidden vector and a backward node hidden vector of the current node in each session graph;
and selecting a recommendation session from the plurality of session sequences according to the vector representation, and recommending to the user.
2. The session recommendation method of claim 1, wherein said generating a vector representation of a current node from a forward node hidden vector and a backward node hidden vector of the current node in each session graph comprises:
the forward node hidden vectors and the backward node hidden vectors are learned bi-directionally using a bi-directional graph neural network to generate the vector representation.
3. The session recommendation method of claim 2, wherein said bi-directionally learning the forward node hidden vectors and backward node hidden vectors using a bi-directional graph neural network to generate the vector representation comprises:
generating an updating function according to the node vector, the weight, the bias parameter and the corresponding out-degree/in-degree matrix of the session graph;
learning the forward node hidden vector and the backward node hidden vector according to the update function to generate the vector representation.
4. The session recommendation method of claim 1, wherein said selecting a recommendation session from said plurality of sequences of sessions according to said vector representation and recommending to a user comprises:
packing the vector representation of all the nodes of each meeting picture to obtain the vector representation of the session corresponding to the current session picture;
scoring the vector representation for each session according to the user's session history and current preferences to select the recommended sessions from the plurality of sequences of sessions for recommendation to the user.
5. A conversation recommendation apparatus, comprising:
the conversation sequence conversion module is used for converting a plurality of conversation sequences into a conversation graph;
a node vector representation generating module, configured to generate a vector representation of a current node according to a forward node hidden vector and a backward node hidden vector of the current node in each session graph;
and the recommendation session selection module is used for selecting a recommendation session from the plurality of session sequences according to the vector representation and recommending the recommendation session to a user.
6. The session recommendation apparatus of claim 5, wherein the node vector representation generation module comprises:
a node vector generation unit configured to bidirectionally learn the forward node hidden vector and the backward node hidden vector using a bi-directional neural network to generate the vector representation.
7. The session recommendation apparatus of claim 6, wherein the node vector generation unit comprises:
the updating function generating unit is used for generating an updating function according to the node vector, the weight, the bias parameter and the corresponding out-degree/in-degree matrix of the session graph;
a node vector representation generating unit, configured to learn the forward node hidden vector and the backward node hidden vector according to the update function to generate the vector representation.
8. The session recommendation device of claim 5, wherein the recommendation session selection module comprises:
the vector representation packing unit is used for packing the vector representations of all the nodes of each meeting picture to obtain the vector representation of the session corresponding to the current session graph;
and the recommendation session selection unit is used for grading the vector representation of each session according to the session history of the user and the current preference so as to select the recommendation session from the plurality of session sequences and recommend the recommendation session to the user.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the steps of the session recommendation method of any one of claims 1 to 4 are implemented when the program is executed by the processor.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the session recommendation method according to any one of claims 1 to 4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110973346.XA CN113656696A (en) | 2021-08-24 | 2021-08-24 | Session recommendation method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110973346.XA CN113656696A (en) | 2021-08-24 | 2021-08-24 | Session recommendation method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113656696A true CN113656696A (en) | 2021-11-16 |
Family
ID=78492626
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110973346.XA Pending CN113656696A (en) | 2021-08-24 | 2021-08-24 | Session recommendation method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113656696A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113961816A (en) * | 2021-11-26 | 2022-01-21 | 重庆理工大学 | Graph convolution neural network session recommendation method based on structure enhancement |
CN114491150A (en) * | 2022-03-28 | 2022-05-13 | 苏州浪潮智能科技有限公司 | Video recommendation method, system, device and computer readable storage medium |
CN114647714A (en) * | 2022-03-30 | 2022-06-21 | 贝壳找房网(北京)信息技术有限公司 | Method and apparatus for assisting dialog |
CN117240657A (en) * | 2023-09-07 | 2023-12-15 | 中国电子产业工程有限公司 | VPN application identification method based on graph matching network |
CN117972219A (en) * | 2024-03-28 | 2024-05-03 | 山东科技大学 | Session recommendation method based on dynamic hypergraph and ranking reconstruction model |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110490717A (en) * | 2019-09-05 | 2019-11-22 | 齐鲁工业大学 | Method of Commodity Recommendation and system based on user conversation and figure convolutional neural networks |
CN111046257A (en) * | 2019-12-09 | 2020-04-21 | 北京百度网讯科技有限公司 | Session recommendation method and device and electronic equipment |
CN112150210A (en) * | 2020-06-19 | 2020-12-29 | 南京理工大学 | Improved neural network recommendation method and system based on GGNN (global warming network) |
CN112364976A (en) * | 2020-10-14 | 2021-02-12 | 南开大学 | User preference prediction method based on session recommendation system |
CN112396492A (en) * | 2020-11-19 | 2021-02-23 | 天津大学 | Conversation recommendation method based on graph attention network and bidirectional long-short term memory network |
US20210173841A1 (en) * | 2019-12-06 | 2021-06-10 | NEC Laboratories Europe GmbH | Answering complex queries in knowledge graphs with bidirectional sequence encoders |
CN112948710A (en) * | 2021-03-22 | 2021-06-11 | 华南师范大学 | Graph neural network-based punishment education recommendation method, system and storage medium |
-
2021
- 2021-08-24 CN CN202110973346.XA patent/CN113656696A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110490717A (en) * | 2019-09-05 | 2019-11-22 | 齐鲁工业大学 | Method of Commodity Recommendation and system based on user conversation and figure convolutional neural networks |
US20210173841A1 (en) * | 2019-12-06 | 2021-06-10 | NEC Laboratories Europe GmbH | Answering complex queries in knowledge graphs with bidirectional sequence encoders |
CN111046257A (en) * | 2019-12-09 | 2020-04-21 | 北京百度网讯科技有限公司 | Session recommendation method and device and electronic equipment |
CN112150210A (en) * | 2020-06-19 | 2020-12-29 | 南京理工大学 | Improved neural network recommendation method and system based on GGNN (global warming network) |
CN112364976A (en) * | 2020-10-14 | 2021-02-12 | 南开大学 | User preference prediction method based on session recommendation system |
CN112396492A (en) * | 2020-11-19 | 2021-02-23 | 天津大学 | Conversation recommendation method based on graph attention network and bidirectional long-short term memory network |
CN112948710A (en) * | 2021-03-22 | 2021-06-11 | 华南师范大学 | Graph neural network-based punishment education recommendation method, system and storage medium |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113961816A (en) * | 2021-11-26 | 2022-01-21 | 重庆理工大学 | Graph convolution neural network session recommendation method based on structure enhancement |
CN114491150A (en) * | 2022-03-28 | 2022-05-13 | 苏州浪潮智能科技有限公司 | Video recommendation method, system, device and computer readable storage medium |
CN114491150B (en) * | 2022-03-28 | 2022-07-15 | 苏州浪潮智能科技有限公司 | Video recommendation method, system, equipment and computer readable storage medium |
CN114647714A (en) * | 2022-03-30 | 2022-06-21 | 贝壳找房网(北京)信息技术有限公司 | Method and apparatus for assisting dialog |
CN117240657A (en) * | 2023-09-07 | 2023-12-15 | 中国电子产业工程有限公司 | VPN application identification method based on graph matching network |
CN117240657B (en) * | 2023-09-07 | 2024-03-12 | 中国电子产业工程有限公司 | VPN application identification method based on graph matching network |
CN117972219A (en) * | 2024-03-28 | 2024-05-03 | 山东科技大学 | Session recommendation method based on dynamic hypergraph and ranking reconstruction model |
CN117972219B (en) * | 2024-03-28 | 2024-06-11 | 山东科技大学 | Session recommendation method based on dynamic hypergraph and ranking reconstruction model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Vesselinova et al. | Learning combinatorial optimization on graphs: A survey with applications to networking | |
CN113656696A (en) | Session recommendation method and device | |
JP7105789B2 (en) | Machine learning method and apparatus for ranking network nodes after using a network with software agents at the network nodes | |
CN110059262B (en) | Project recommendation model construction method and device based on hybrid neural network and project recommendation method | |
Bergstra et al. | Hyperopt: a python library for model selection and hyperparameter optimization | |
CN111291266A (en) | Artificial intelligence based recommendation method and device, electronic equipment and storage medium | |
US20190236479A1 (en) | Method and apparatus for providing efficient testing of systems by using artificial intelligence tools | |
Gibert et al. | Choosing the right data mining technique: classification of methods and intelligent recommendation | |
US20220414470A1 (en) | Multi-Task Attention Based Recurrent Neural Networks for Efficient Representation Learning | |
WO2023163774A1 (en) | Individual treatment effect estimation under high-order interference in hypergraphs taking into account spillover effects | |
CN114595383A (en) | Marine environment data recommendation method and system based on session sequence | |
Firouzian et al. | Investigation of the effect of concept drift on data-aware remaining time prediction of business processes | |
US20220044136A1 (en) | Automated data table discovery for automated machine learning | |
Dharmasena et al. | Modeling mobile apps user behavior using Bayesian networks | |
Roa et al. | A knowledge-based equation discovery system for engineering domains | |
CN113569130A (en) | Content recommendation method, device, equipment and readable storage medium | |
Attya et al. | Novel framework for selecting cloud provider using neutrosophic and modified gan | |
Makarova et al. | A case-based reasoning approach with fuzzy linguistic rules: Accuracy validation and application in interface design-support intelligent system | |
Mohammed et al. | Location-aware deep learning-based framework for optimizing cloud consumer quality of service-based service composition | |
Štolfa et al. | Value estimation of the use case parameters using SOM and fuzzy rules | |
Mattyasovszky-Philipp et al. | Adaptive/cognitive resonance and the architecture issues of cognitive information systems | |
Nguyen et al. | DEEP LEARNING FOR SIMULTANEOUS IMPUTATION AND CLASSIFICATION OF TIME SERIES INCOMPLETE DATA | |
Chung et al. | Introduction to artificial intelligence (AI): Definition and scope of AI | |
Susanti et al. | Link Prediction in Educational Graph Data to Predict Elective Course using Graph Convolutional Network Model | |
EP4235505A1 (en) | System for probabilistic reasoning and decision making on digital twins |
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 |