CN112044082A - Information detection method and device and computer readable storage medium - Google Patents
Information detection method and device and computer readable storage medium Download PDFInfo
- Publication number
- CN112044082A CN112044082A CN202010884149.6A CN202010884149A CN112044082A CN 112044082 A CN112044082 A CN 112044082A CN 202010884149 A CN202010884149 A CN 202010884149A CN 112044082 A CN112044082 A CN 112044082A
- Authority
- CN
- China
- Prior art keywords
- matrix
- target
- information
- vector matrix
- detected
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63F—CARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
- A63F13/00—Video games, i.e. games using an electronically generated display having two or more dimensions
- A63F13/70—Game security or game management aspects
- A63F13/79—Game security or game management aspects involving player-related data, e.g. identities, accounts, preferences or play histories
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63F—CARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
- A63F13/00—Video games, i.e. games using an electronically generated display having two or more dimensions
- A63F13/70—Game security or game management aspects
- A63F13/75—Enforcing rules, e.g. detecting foul play or generating lists of cheating players
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63F—CARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
- A63F13/00—Video games, i.e. games using an electronically generated display having two or more dimensions
- A63F13/80—Special adaptations for executing a specific game genre or game mode
- A63F13/837—Shooting of targets
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63F—CARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
- A63F2300/00—Features of games using an electronically generated display having two or more dimensions, e.g. on a television screen, showing representations related to the game
- A63F2300/80—Features of games using an electronically generated display having two or more dimensions, e.g. on a television screen, showing representations related to the game specially adapted for executing a specific type of game
- A63F2300/8076—Shooting
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Business, Economics & Management (AREA)
- Computer Security & Cryptography (AREA)
- General Business, Economics & Management (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the application discloses an information detection method, an information detection device and a computer readable storage medium, wherein a target incidence matrix formed by acquiring a characteristic matrix and an interactive behavior of an object set to be detected is obtained; transmitting information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the feature matrix to obtain a target object representation vector matrix; generating label information corresponding to the target object representation vector matrix; training the preset neural network model through the target object representation vector matrix and the corresponding label information to obtain the trained preset neural network model; and carrying out identity detection on the target object to be recognized based on the trained preset neural network model. Therefore, intimacy information implicit in the target incidence matrix is transmitted to the feature matrix, label calibration and training are carried out, the trained preset neural network model can be combined with the neighbor object relation in the object set to carry out identity recognition, and the accuracy of information detection is greatly improved.
Description
Technical Field
The present application relates to the field of communications technologies, and in particular, to an information detection method, an information detection apparatus, and a computer-readable storage medium.
Background
With the rapid development of internet technology, the processing capability of an intelligent terminal processor is stronger and stronger, and many game applications are derived, such as online First-person shooter games (FPSs), which can support a plurality of players to be online simultaneously and perform vivid interactive entertainment based on rich scenes.
In the prior art, in order to prevent the minors from being excessively enthusiastic in the game, the minor players need to be identified and limited, the minor players can be identified by performing real-name authentication on the game players in the practical application process, and the game authority is not given to the players without the real-name authentication.
In the process of research and practice of the prior art, the inventor of the application finds that in the prior art, the underage player can perform real-name verification by falsely using the identity of other people, so that the result of information detection is inaccurate.
Disclosure of Invention
The embodiment of the application provides an information detection method, an information detection device and a computer-readable storage medium, which can further improve the accuracy of information detection.
In order to solve the above technical problem, an embodiment of the present application provides the following technical solutions:
an information detection method, comprising:
acquiring a characteristic matrix of an object set to be detected and a target incidence matrix formed by interactive behaviors;
transmitting the information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the feature matrix to obtain a target object representation vector matrix;
generating label information corresponding to the target object representation vector matrix;
training a preset neural network model through the target object representation vector matrix and corresponding label information to obtain the trained preset neural network model;
and carrying out identity detection on the target object to be recognized based on the trained preset neural network model.
An information detection apparatus comprising:
the acquisition unit is used for acquiring a characteristic matrix of the set of objects to be detected and a target incidence matrix formed by interactive behaviors;
the propagation unit is used for propagating the compactness information which characterizes different objects to be detected in the set to be detected in the target incidence matrix to the characteristic matrix to obtain a target object characterization vector matrix;
the generating unit is used for generating label information corresponding to the target object representation vector matrix;
the training unit is used for training a preset neural network model through the target object representation vector matrix and the corresponding label information to obtain the trained preset neural network model;
and the detection unit is used for carrying out identity detection on the target object to be recognized based on the trained preset neural network model.
In some embodiments, the training unit is to:
inputting the object characterization vector information of each line into the preset neural network model, and outputting corresponding prediction information;
inputting the prediction information and the corresponding label information into a loss function;
and carrying out iterative processing on the network parameters of the preset neural network model according to the loss values calculated by the loss function until convergence, so as to obtain the trained preset neural network model.
In some embodiments, the detection unit is configured to:
acquiring a characteristic matrix to be recognized of an object set to be recognized and a target association matrix to be recognized formed by interactive behaviors;
transmitting the information of the compactness between different objects to be detected in the set to be detected in the incidence matrix of the object to be recognized to the characteristic matrix to be recognized to obtain a characteristic vector matrix of the object to be recognized;
and inputting the representation vector matrix of the target object to be recognized into the trained neural network model, and outputting the identity detection result of each target object to be recognized in the set of the target objects to be recognized.
The method comprises the steps of obtaining a characteristic matrix of an object set to be detected and a target incidence matrix formed by interactive behaviors; transmitting information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the feature matrix to obtain a target object representation vector matrix; generating label information corresponding to the target object representation vector matrix; training the preset neural network model through the target object representation vector matrix and the corresponding label information to obtain the trained preset neural network model; and carrying out identity detection on the target object to be recognized based on the trained preset neural network model. Therefore, intimacy information implicit in the target incidence matrix is transmitted to the feature matrix, label calibration and training are carried out, the trained preset neural network model can be combined with the neighbor object relation in the object set to carry out identity recognition, and the accuracy of information detection is greatly improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic view of a scene of an information detection system provided in an embodiment of the present application;
fig. 2 is a schematic flowchart of an information detection method provided in an embodiment of the present application;
fig. 3 is a schematic flow chart of an information detection method according to an embodiment of the present application;
FIG. 4 is a schematic structural diagram of an information detection apparatus provided in an embodiment of the present application;
fig. 5 is a schematic structural diagram of a server according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. 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 application.
The embodiment of the application provides an information detection method, an information detection device and a computer readable storage medium.
Referring to fig. 1, fig. 1 is a schematic view of a scene of an information detection system according to an embodiment of the present application, including: the terminal a and the server (the information detection system may also include other terminals besides the terminal a, and the specific number of the terminals is not limited herein), the terminal a and the server may be connected through a communication network, which may include a wireless network and a wired network, wherein the wireless network includes one or more of a wireless wide area network, a wireless local area network, a wireless metropolitan area network, and a wireless personal area network. The network includes network entities such as routers, gateways, etc., which are not shown in the figure. The terminal a may perform information interaction with the server through the communication network, for example, the terminal a may send the target object to be identified, which needs to be subjected to identity detection, to the server.
The information detection system can comprise an information detection device, the information detection device can be specifically integrated in a server, the server can be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and can also be a cloud server for providing basic cloud computing services such as cloud service, a cloud database, cloud computing, cloud functions, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN (content delivery network) and big data and artificial intelligence platforms. As shown in fig. 1, the server may obtain a feature matrix of a set of objects to be detected and a target association matrix formed by interaction behaviors; transmitting the information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the characteristic matrix to obtain a target object representation vector matrix; generating label information corresponding to the target object representation vector matrix; training a preset neural network model through the target object representation vector matrix and corresponding label information to obtain the trained preset neural network model; and receiving the target object to be recognized sent by the terminal A, carrying out identity detection on the target object to be recognized based on the trained preset neural network model, generating an identity detection result and returning the identity detection result to the terminal A.
The terminal A in the information detection system can be provided with various applications required by users, such as instant messaging application, media application, game application and the like, and can send a target object to be identified to the server for identity detection based on the game application and receive an identity detection result returned by the server.
It should be noted that the scene schematic diagram of the information detection system shown in fig. 1 is only an example, and the information detection system and the scene described in the embodiment of the present application are for more clearly illustrating the technical solution of the embodiment of the present application, and do not form a limitation on the technical solution provided in the embodiment of the present application.
The following are detailed below. The numbers in the following examples are not intended to limit the order of preference of the examples.
In the present embodiment, the information detection apparatus will be described in terms of an information detection apparatus, which may be specifically integrated into a computer device having a storage unit and a microprocessor installed therein and having an arithmetic capability, the computer device may be a server or a terminal, and the computer device is exemplified as a server in the present embodiment.
Referring to fig. 2, fig. 2 is a schematic flow chart of an information detection method according to an embodiment of the present disclosure. The information detection method comprises the following steps:
in step 101, a feature matrix of an object set to be detected and a target association matrix formed by interactive behaviors are obtained.
Artificial Intelligence (AI) is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human Intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making.
The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.
Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and formal education learning.
The scheme provided by the embodiment of the application relates to the technologies such as the machine learning technology of artificial intelligence and the like, and is specifically explained by the following embodiment:
it should be noted that, with the popularity of game applications, consumption and game behaviors of minors need to be limited, so that authentication of game objects (i.e. game users) in a game is required to determine whether the game objects are adult game objects or immature game objects, and in the prior art, immature players can be identified through real-name authentication, that is, the game objects are adult players or immature players are determined according to identity card information by collecting identity card information of the game objects, but in the actual application process, the immature players can use identities of other people to perform real-name authentication, so that authentication failure often occurs.
Among them, in an actual game, although minor persons can evade authentication of the system by impersonating others, some fixed characteristics are determined due to the game behavior of minor game objects, for example, minor game objects mainly play a voluntary team game with persons of similar ages, and game time, game operations, and game consumption of minor game objects are significantly different from those of the major game objects.
In the embodiment of the present application, the object set to be detected is composed of a plurality of game objects forming a group voluntarily, based on the above logic rules, a feature matrix of the object set to be detected may be obtained, where the feature matrix may be composed of a number dimension of the object to be detected and a feature dimension of the game object, the feature of the game object may include game formation features, game behavior features (i.e., operation level, game score, and the like), game consumption features, game time features, and the like, and the feature matrix reflects game behavior features of each game object.
Further, a target association matrix formed by interaction between different to-be-detected objects in the to-be-detected object set can be obtained, for example, when a presentation behavior occurs between every two to-be-detected objects, the presentation behaviors of the two to-be-detected objects can be recorded, so that a matrix for carrying out presentation behavior interaction between different to-be-detected objects is formed, the target association matrix can be formed by combining a plurality of association matrices formed by interaction behaviors of the to-be-detected object set in different dimensions, and can reflect the association between different to-be-detected objects in the to-be-detected object set, and the matrix length of the number dimension of the to-be-detected objects and the matrix width of the number dimension of the to-be-detected objects reflect the interaction behavior rules of each to-be-detected object and other to-be.
In some embodiments, the step of obtaining a feature matrix of the set of objects to be detected and a target association matrix formed by the interaction behavior may include:
(1) acquiring a characteristic matrix of an object set to be detected;
(2) acquiring a plurality of incidence matrixes formed by the interactive behaviors of an object set to be detected under different feature dimensions;
(3) performing matrix standardization on each incidence matrix to obtain a plurality of incidence matrices subjected to matrix standardization;
(4) and carrying out weighted aggregation on the plurality of incidence matrixes subjected to matrix standardization to obtain the target incidence matrix.
The feature matrix of the set of objects to be detected can be obtained, the feature matrix can be represented as X, the dimension of X is N × F, the N is the number dimension of the objects to be detected in the set of objects to be detected, the F represents the feature dimension of the objects to be detected, the feature dimension can include a plurality of features, for example, the features include game formation features, game chat times features and game presentation features, as the underage game objects mainly play voluntary formation games with similar ages, and the game formation behaviors, game chat behaviors and game consumption average behaviors of the underage game objects are significantly different from the adult game objects, the game formation features, the game chat times features and the game presentation features can be obtained to reflect the ages of users.
Further, a plurality of incidence matrixes formed by the interaction behaviors of the object set to be detected under different feature dimensions are obtained, for example, a first incidence matrix formed by the interaction of the object set to be detected in game team feature dimensions is obtained, the first incidence matrix is N x N, the N is the number dimension of the objects to be detected in the object set to be detected, and represents the team interaction behavior representation of each object to be detected and other objects to be detected in each object set to be detected. By analogy, a second association matrix N x N representing the chat interaction behavior representation of each object to be detected and other objects to be detected in each set of objects to be detected and a third association matrix N x N representing the game presentation behavior representation can be continuously obtained.
In order to better combine the matrix information in different dimensions, in the embodiment of the present application, by performing a matrix normalization process on each correlation matrix, that is, the elements in each incidence matrix are classified as (0, 1), so that the subsequent uniform processing can be performed, and when a plurality of incidence matrices after the matrix standardization processing are obtained, the plurality of correlation matrices after the matrix normalization can be weighted and aggregated, and the plurality of correlation matrices after each matrix normalization are weighted, because the length and the width of the matrices are the same, after weighting, the elements of the plurality of correlation matrices may be correspondingly added according to the row-column arrangement rule to combine into a target correlation matrix, the target incidence matrix may be the incidence of different objects to be detected in the set of objects to be detected under the condition of integrating all characteristic dimensions, and reflects the closeness relation of all the objects to be detected in the set of objects to be detected.
In step 102, the information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix is transmitted to the feature matrix, so as to obtain a target object representation vector matrix.
In the embodiment of the application, each object to be detected in the object set to be detected can be regarded as a node in the graph neural network, and each object to be detected can obtain the characteristic information of the object to be detected of the neighbor node in the process of graph propagation, so that each object to be detected can contain more layers of characteristic information, and the actual behavior characteristics of the team are fused.
Therefore, by the principle, a graph structure is constructed for the target incidence matrix and the feature matrix, graph propagation is continuously carried out through a graph neural network, the information representing the compactness between different objects to be detected in a set to be detected in the target incidence matrix is propagated to the feature matrix, the compactness information represents the incidence relation between each object to be detected and the feature information of the neighbor objects to be detected, a target object representation vector matrix is obtained, the target object representation vector matrix reflects the feature information of each object to be detected and the incidence relation between each object to be detected and other objects to be detected, behaviors representing game objects are fused from multiple dimensions, representation behaviors of each user to be detected after the information transmitted by interactive neighbor nodes is combined can be reflected, and a plurality of similar behavior features are bound among different users in an immature team, the behavior characteristics are mutually transmitted, a target object representation vector matrix representing the identity of each user to be detected can be obtained, and the identity of the game object can be more accurately defined.
In some embodiments, the step of transmitting the information representing the closeness between different objects to be detected in the set to be detected in the target association matrix to the feature matrix to obtain the target object representation vector matrix may include:
(1) determining the target incidence matrix as an object characterization vector matrix;
(2) inputting the feature matrix and the object characterization vector matrix into a propagation function of a neural network of the graph;
(3) matrix multiplication processing, linear transformation processing and nonlinear activation processing are carried out on the feature matrix and the object characterization vector matrix to obtain an updated object characterization vector matrix;
(4) and repeatedly executing preset times to perform matrix multiplication processing, linear transformation processing and nonlinear activation processing on the feature matrix and the updated object representation vector matrix to obtain an updated object representation vector matrix, thereby obtaining the target object representation vector matrix.
In which, the propagation function of the figure neural network can be referred to together:
the Hl+1An object characterization vector matrix representing the next layer (updated object characterization vector matrix), which should be the activation function,is a target incidence matrix, HlCharacterizing a vector matrix for the current object, the initial object characterizing vector momentsThe array being a feature matrix, WlParameters are propagated for the network of the current layer.
Therefore, the feature matrix is determined as an object characterization vector matrix through the formula, the target incidence matrix and the object characterization vector matrix are input into the propagation function of the graph neural network, matrix multiplication processing is carried out on the target incidence matrix and the object characterization vector matrix, feature information of neighbor nodes is absorbed, linear transformation is carried out on the result of the matrix multiplication processing and network propagation parameters of the current layer, and then nonlinear processing is carried out through an activation function, so that the object characterization vector matrix of the next layer is obtained.
Further, in order to acquire more feature information of the neighbor nodes for absorption as much as possible, propagation for a preset number of times is required, that is, the preset number of times is repeatedly executed to perform matrix multiplication processing, linear transformation processing and nonlinear activation processing on the target association matrix and the updated object characterization vector, so as to obtain an updated object characterization vector matrix, and obtain a propagated target object characterization vector matrix.
In some embodiments, the propagation function of the graph neural network further includes object characterization recall coefficients and propagation parameter recall coefficients, and the step of performing matrix multiplication processing, linear transformation processing and nonlinear activation processing on the target correlation matrix and the object characterization vector matrix to obtain an updated object characterization vector matrix may include:
(1.1) carrying out matrix multiplication processing and linear transformation processing on the target incidence matrix and the object characterization vector matrix, supplementing gradient information processing and nonlinear activation processing of an initial object characterization vector matrix according to an object characterization recall coefficient and a propagation parameter recall coefficient, and obtaining an updated object characterization vector matrix.
In order to avoid that the characteristic information before the graph neural network forgets after long-distance propagation and affects the accuracy of subsequent matrix processing, a user characterization recall coefficient and a propagation parameter recall coefficient may be introduced in the embodiment of the present application, so that after matrix multiplication processing and line transformation processing are performed on the target association matrix and the object characterization vector matrix, gradient information of an initial object characterization vector matrix with a certain weight may be supplemented according to the object characterization recall coefficient and the propagation parameter recall coefficient, so that no matter how many times propagation function processing is performed, residual information of the initial object characterization vector matrix is retained, matrix data redundancy is avoided, and nonlinear activation processing is performed after supplementation, so as to obtain an updated object characterization vector matrix.
In some embodiments, the repeatedly performing a preset number of times to perform matrix multiplication processing, linear transformation processing, and nonlinear activation processing on the target correlation matrix and the updated object token vector matrix to obtain an updated object token vector matrix, and the obtaining a target object token vector matrix includes:
and (2.2) repeatedly executing preset times to perform matrix multiplication processing and linear transformation processing on the target incidence matrix and the updated object characterization vector matrix, supplementing gradient information processing and nonlinear activation processing of the initial object characterization vector matrix according to the object characterization recall coefficient and the propagation parameter recall coefficient, and obtaining the updated object characterization vector matrix to obtain the target object characterization vector matrix.
The method comprises the steps of repeatedly executing preset times to perform matrix multiplication processing, linear transformation processing, gradient information processing and nonlinear activation processing on the target association matrix and the updated object characterization vector matrix according to an object characterization recall coefficient and a propagation parameter recall coefficient to supplement an initial object characterization vector matrix to obtain an updated object characterization vector matrix, and obtaining the target object characterization vector matrix, wherein the target object characterization vector matrix can be N x 50, N is the number dimension of objects to be detected in an object set to be detected, and as the object characterization vector matrix information is continuously supplemented with the gradient information processing of the initial object characterization vector matrix according to the object characterization recall coefficient and the propagation parameter recall coefficient in the propagation process, relative to the graph neural network propagation function in the related technology, the initial object characterization recall coefficient and the propagation parameter recall coefficient can be continuously kept by the object characterization recall coefficient and the propagation parameter recall coefficient Original representation information of the matrix can not lose the original representation relation due to supplement of neighbor information, and the representation information of the original object representation vector matrix is reserved, so that the representation of the characteristics is more accurate.
In step 103, label information corresponding to the target object representation vector matrix is generated.
The target object representation vector matrix comprises representation relations of relations between N objects to be detected and other objects to be detected, the objects to be detected can be represented in a fusion mode from multiple dimensions, the compactness of the objects to be detected and the compactness of the objects to be detected in the set of the objects to be detected are rich in representation capacity, accordingly, the graph propagation characteristics (namely object representation vector information) of each object to be detected in the target object representation vector matrix can be obtained, corresponding label information is generated according to actual conditions, for example, when the objects to be detected are minor player objects, a label 1 is generated, and when the objects to be detected are adult player objects, a label 0 is generated.
The target object representation vector matrix can reflect the interactive representation behaviors of each user to be detected after the information transmitted by the interactive neighbor nodes is combined, and different users in an underage team tend to have many similar behavior characteristics, the behavior characteristics are mutually transmitted and refined, and the identity is marked for training, so that the identity of the user can be judged according to the behaviors by a model.
In some embodiments, the step of generating tag information corresponding to the target object characterization vector matrix may include:
(1) sequentially acquiring object representation vector information of each row in the target object representation vector matrix;
(2) tag information of the object characterization vector information of each line is generated.
The number of rows of the target object representation vector matrix is the number N of the objects to be detected in the object set to be detected, so that the object representation vector information of each row in the target object representation vector matrix can be sequentially acquired, and the object representation vector information of each row represents the image propagation characteristics of each object to be detected.
Further, the label information is calibrated according to whether the object to be detected in each row is an adult game object or an immature game object, for example, when the object to be detected is an immature player object, a label 1 is generated, and when the object to be detected is an adult player object, a label 0 is generated.
In step 104, the preset neural network model is trained through the target object representation vector matrix and the corresponding label information, so as to obtain the trained preset neural network model.
The preset neural network model can be a multilayer perceptron (MLP), the graph propagation characteristics of each object to be detected in the target object characterization vector matrix and the corresponding calibrated label information can be input into the preset neural network model for training, the preset neural network model performs continuous characteristic extraction on the graph propagation characteristics of each object to be detected, final prediction information is output, network parameters of the preset neural network model are continuously adjusted according to the difference between the prediction information and the corresponding label information until the difference is infinitesimal, and the trained preset neural network model is obtained.
In some embodiments, the training the preset neural network model through the target object characterization vector matrix and the corresponding label information to obtain the trained preset neural network model may include:
(1) inputting the object characterization vector information of each line into the preset neural network model, and outputting corresponding prediction information;
(2) inputting the prediction information and the corresponding label information into a loss function;
(3) and carrying out iterative processing on the network parameters of the preset neural network model according to the loss values calculated by the loss function until convergence to obtain the trained preset neural network model.
The method comprises the steps of inputting object characterization vector information of each row in a target object characterization vector matrix into a preset neural network model for feature extraction, outputting corresponding prediction information, inputting the prediction information and corresponding label information into a loss function, wherein the loss function is used for calculating a value (namely a loss value) of a square difference between the corresponding label information and the prediction information, and in the early stage, because network parameters are randomly selected, the loss value is large, the loss value can be propagated reversely according to the loss value calculated by the loss function, network parameters of the preset neural network model are adjusted and iteratively adjusted until the loss value is converged, and the preset neural network model with more accurate prediction is obtained.
In step 105, identity detection is performed on the target object to be recognized based on the trained preset neural network model.
When the target object to be recognized is voluntarily grouped in the game, all the objects to be recognized in the group can be collected to form an object set to be recognized, a feature matrix of the object set to be recognized and a target incidence matrix to be recognized formed by interaction behaviors are further obtained, information representing the compactness between different objects to be detected in the set to be recognized in the target incidence matrix to be recognized is transmitted to the feature matrix to be recognized, and a target object representation vector matrix to be recognized is obtained.
Furthermore, target object characterization vector information corresponding to the target object to be recognized is taken from the target object characterization vector matrix to be recognized and input into the trained preset neural network model, the training of the preset neural network model in the early stage learns the capability of recognizing the target object characterization vector information, so that prediction information can be output, the prediction information belongs to the range between (0 and 1), when the prediction information is larger than 0.5, the target object to be recognized is judged to be an underage player, when the prediction information is not larger than 0.5, the target object to be recognized is judged to be an adult player, and therefore, the target object to be recognized can be automatically subjected to identity detection according to game operation behavior rules of the underage player and the adult player in a team, real-name authentication is carried out without depending on the identity card information of the user, and non-sensitive identity detection is realized.
In some embodiments, the step of performing identity detection on the target object to be recognized based on the trained preset neural network model may include:
(1) acquiring a characteristic matrix to be recognized of an object set to be recognized and a target association matrix to be recognized formed by interactive behaviors;
(2) transmitting the information of the compactness between different objects to be detected in the set to be detected in the incidence matrix of the object to be recognized to the characteristic matrix to be recognized to obtain a characteristic vector matrix of the object to be recognized;
(3) and inputting the representation vector matrix of the target object to be recognized into the trained neural network model, and outputting the identity detection result of each target object to be recognized in the set of the target objects to be recognized.
When a game player autonomously forms a team, a set of objects to be recognized formed by all objects to be recognized in the team is collected, a feature matrix to be recognized of the set of objects to be recognized and a target association matrix to be recognized formed by mutual behaviors are collected in real time, information representing the compactness between different objects to be detected in the set to be recognized in the target association matrix to be recognized is transmitted to the feature matrix to be recognized, and a target object representation vector matrix to be recognized is obtained.
Further, the representation vector information of the object to be recognized in each row in the representation vector matrix of the object to be recognized is sequentially input into the trained neural network model, the prediction information (prediction score) of each object to be recognized in the set of objects to be recognized is output, when the prediction information is greater than 0.5, the object to be recognized is determined to be an immature player, and when the prediction information is not greater than 0.5, the object to be recognized is determined to be an adult player.
Therefore, the target association matrix formed by the characteristic matrix and the interactive behavior of the set of the objects to be detected is obtained; transmitting information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the feature matrix to obtain a target object representation vector matrix; generating label information corresponding to the target object representation vector matrix; training the preset neural network model through the target object representation vector matrix and the corresponding label information to obtain the trained preset neural network model; and carrying out identity detection on the target object to be recognized based on the trained preset neural network model. Therefore, intimacy information implicit in the target incidence matrix is transmitted to the feature matrix, label calibration and training are carried out, the trained preset neural network model can be combined with the neighbor object relation in the object set to carry out identity recognition, and the accuracy of information detection is greatly improved.
The method described in connection with the above embodiments will be described in further detail below by way of example.
In the present embodiment, the information detection apparatus will be described by taking an example in which it is specifically integrated in a server.
Referring to fig. 3, fig. 3 is another schematic flow chart of an information detection method according to an embodiment of the present application, where the method flow may include:
in step 201, the server obtains a feature matrix of the set of objects to be detected, and obtains a plurality of association matrices formed by interaction behaviors of the set of objects to be detected in different feature dimensions.
The server obtains a feature matrix X of the set of objects to be detected, where a dimension of X is N × F, where N is a number dimension of the objects to be detected in the set of objects to be detected, and F represents a feature dimension of the objects to be detected, where the feature dimension may include multiple features, such as a game team feature, a game chat number feature, a game presentation feature, a game time feature, a game behavior feature, a game consumption feature, and the like.
Further, a plurality of association matrixes formed by the interaction behaviors of the set of objects to be detected under different feature dimensions are obtained, and if 3 dimension features capable of having the interaction behaviors are assumed, the dimension features are respectively game team features, game chat time features and game presentation features, then 3 association matrixes for the behavior interaction between each object to be detected and other objects to be detected under the game team features, the game chat time features and the game presentation features can be collected, and all the association matrixes are N.
In step 202, the server performs matrix normalization on each incidence matrix to obtain a plurality of incidence matrices after the matrix normalization, and performs weighted aggregation on the plurality of incidence matrices after the matrix normalization to obtain a target incidence matrix.
In order to combine 3 different incidence matrices, the server may perform matrix normalization processing through a matrix normalization formula, which refers to the following formula:
d is a Laplace matrix, the elements except the diagonal are all zero elements, I is an identity matrix, and therefore, matrix normalization processing is carried out on each incidence matrix through the formula to obtain a plurality of incidence matrixes after matrix normalization processing, the elements in the incidence matrixes after matrix normalization processing are all numerical values between (0, 1), and the value is shown hereAnd normalizing the processed correlation matrixes.
Further, in order to reduce the computational complexity, the elements in the multiple incidence matrices after the matrix normalization process may be weighted and aggregated by a dynamic weighting and aggregating matrix, see the following formula:
the wkFor the weight matrix, K represents the normalized correlation matrix of the Kth matrix, and wkThe method is used for weighting the incidence matrix after matrix standardization processing to avoid the situation of overlarge elements, and summing the incidence matrix after weighting processing to obtain a target incidence matrix, whereinIs the target incidence matrix.
In step 203, the server determines the feature matrix as an object characterization vector matrix, inputs the target incidence matrix and the object characterization vector matrix into a propagation function of the neural network, and performs matrix multiplication processing, linear transformation processing, gradient information processing and nonlinear activation processing for supplementing the initial object characterization vector matrix according to the object characterization recall coefficient and the propagation parameter recall coefficient to obtain an updated object characterization vector matrix.
The propagation function of the neural network further includes object representation recall coefficients and propagation parameter recall coefficients, please refer to the following formula:
the Hl+1An object characterization vector matrix representing the next layer (updated object characterization vector matrix), which should be the activation function,is a target incidence matrix, HlThe vector matrix is characterized for the current object, and the initial object characterization vector matrix is the feature matrix X (i.e. H)0X), theIs the network propagation parameter of the current layer, the alphalThe recall coefficient for the object may be 0.1, beta1For propagation parameter recall coefficients, may be 0.9,the network propagation parameters for the initial layer.
Therefore, the server can determine the feature matrix X as an initial object characterization vector matrix and associate a target with the matrixInputting the propagation function of the neural network, sequentially performing matrix multiplication on the target incidence matrix and the object characterization vector matrix, and performing matrix multiplication on the target incidence matrix and the object characterization vector matrixAnd transmitting the intimacy feature description of the adjacent relation to an object characterization vector matrix, and then multiplying the result of the matrix by the network transmission parameter of the current layerAnd linear transformation is carried out, so that in order to avoid that the processing result is greatly shifted because the propagation function of the surface map neural network forgets the previous characteristic information after long-distance propagation, the gradient information of the initial object characterization vector matrix can be supplemented according to the object characterization recall coefficient and the propagation parameter recall coefficient, and finally, nonlinear processing is carried out through an activation function sigma to obtain the next layer of object characterization vector matrix (updated object characterization vector matrix).
In step 204, the server repeatedly executes preset times to perform matrix multiplication processing, linear transformation processing, gradient information processing and nonlinear activation processing for supplementing the initial object representation vector matrix according to the object representation recall coefficient and the propagation parameter recall coefficient on the target association matrix and the updated object representation vector matrix, so as to obtain an updated object representation vector matrix, and obtain the target object representation vector matrix.
The server can repeatedly execute the steps of matrix multiplication processing, linear transformation processing, gradient information processing and nonlinear activation processing which supplement the initial object representation vector matrix according to the object representation recall coefficient and the propagation parameter recall coefficient for preset times to the target incidence matrix and the updated object representation vector matrix to obtain the updated object representation vector matrix, so that H isl+1Continuously updating, continuously learning the intimacy relationship between the objects to be detected in each updating, and continuously conducting the characteristic information of the neighbor nodes with each other, assuming that the preset times are 10 timesAfter conducting for 10 times, a target object characterization vector matrix is obtained, which can not only characterize the feature expression of each object to be detected, but also combine the intimacy feature expression of the adjacent objects to be detected, because the immature game objects mainly conduct voluntary team formation games with similar age people, and the game time, game operation and game consumption of the immature game objects are all significantly different from those of the adult game objects, so that the intimacy feature expressions of the adjacent objects to be detected are combined, the feature expression capability of the target object characterization vector matrix for identity can be better, the target object characterization vector matrix can be N x 50, where N is the number dimension of the objects to be detected in the set of objects to be detected, and 50 is the number of the feature dimensions after propagation.
In step 205, the server sequentially obtains object representation vector information of each row in the target object representation vector matrix, and generates tag information of the object representation vector information of each row.
The server may sequentially obtain object characterization vector information of each row in the target object characterization vector matrix, that is, sequentially obtain object characterization vector information of each object to be detected, and perform label calibration according to whether each object to be detected is an adult game player or an immature game player, label information 1 is calibrated for the object characterization vector information of the object to be detected if the object to be detected is an adult game player, and label information 0 is calibrated for the object characterization vector information of the object to be detected if the object to be detected is an adult game player.
In step 206, the server inputs the object characterization vector information of each row into the preset neural network model, outputs corresponding prediction information, inputs the prediction information and corresponding label information into a loss function, and performs iterative processing on network parameters of the preset neural network model according to a loss value calculated by the loss function until convergence to obtain the trained preset neural network model.
Inputting object characterization vector information of each object to be detected into the preset neural network model for feature extraction, outputting corresponding prediction information, and inputting the prediction information and corresponding label information into a loss function, wherein the loss function refers to the following formula:
the Y isiLabel information representing an object to be detected, thePrediction information representing an object to be detected, LiAnd representing the loss value, inputting the prediction information and the corresponding label information into a loss function based on the formula, and performing iterative processing on the network parameters of the preset neural network model according to the loss value until the loss value is converged to obtain the trained preset neural network model.
In step 207, the server obtains a feature matrix to be recognized of the set of objects to be recognized and a target association matrix to be recognized formed by the interaction behavior.
In step 208, the server transmits the information representing the closeness between different objects to be detected in the set to be detected in the target association matrix to be identified to the feature matrix to be identified, so as to obtain a target object representation vector matrix to be identified.
When the server detects that the target object to be recognized performs voluntary team formation in the game, the server can collect all the objects to be recognized in the team to form a set of objects to be recognized, further obtain a feature matrix of the set of objects to be recognized and a target association matrix to be recognized formed by interaction behaviors, and transmit information representing the closeness between different objects to be detected in the set to be recognized in the target association matrix to be recognized to the feature matrix to be recognized to obtain a target object representation vector matrix to be recognized.
In step 209, the server inputs the characterization vector matrix of the target object to be recognized into the trained neural network model, and outputs the identity detection result of each target object to be recognized in the set of objects to be recognized.
The server inputs the representation vector information of the target object to be recognized in each row in the representation vector matrix of the target object to be recognized into the trained neural network model in sequence, outputs the prediction information (prediction score) of each target object to be recognized in the set of the target object to be recognized, judges that the target object to be recognized is the identity detection result of the non-adult player when the prediction information is greater than 0.5, and judges that the target object to be recognized is the identity detection result of the adult player when the prediction information is not greater than 0.5.
Therefore, the target association matrix formed by the characteristic matrix and the interactive behavior of the set of the objects to be detected is obtained; transmitting information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the feature matrix to obtain a target object representation vector matrix; generating label information corresponding to the target object representation vector matrix; training the preset neural network model through the target object representation vector matrix and the corresponding label information to obtain the trained preset neural network model; and carrying out identity detection on the target object to be recognized based on the trained preset neural network model. Therefore, intimacy information implicit in the target incidence matrix is transmitted to the feature matrix, label calibration and training are carried out, the trained preset neural network model can be combined with the neighbor object relation in the object set to carry out identity recognition, and the accuracy of information detection is greatly improved.
In order to better implement the information detection method provided by the embodiment of the present application, the embodiment of the present application further provides a device based on the information detection method. The terms are the same as those in the above information detection method, and specific implementation details can be referred to the description in the method embodiment.
Referring to fig. 4, fig. 4 is a schematic structural diagram of an information detection apparatus according to an embodiment of the present disclosure, where the information detection apparatus may include an obtaining unit 301, a propagation unit 302, a generation unit 303, a training unit 304, a detection unit 305, and the like.
The obtaining unit 301 is configured to obtain a feature matrix of the set of objects to be detected and a target association matrix formed by the interaction behavior.
In some embodiments, the obtaining unit 301 is configured to:
acquiring a characteristic matrix of an object set to be detected;
acquiring a plurality of incidence matrixes formed by the interactive behaviors of an object set to be detected under different feature dimensions;
performing matrix standardization on each incidence matrix to obtain a plurality of incidence matrices subjected to matrix standardization;
and carrying out weighted aggregation on the plurality of incidence matrixes subjected to matrix standardization to obtain the target incidence matrix.
And the propagation unit 302 is configured to propagate the information, which is used for characterizing the compactness between different objects to be detected in the set to be detected, in the target association matrix to the feature matrix, so as to obtain a target object characterization vector matrix.
In some embodiments, the propagation unit 302 includes:
the determining subunit is used for determining the feature matrix as an object characterization vector matrix;
the input subunit is used for inputting the target incidence matrix and the object characterization vector matrix into a propagation function of the neural network of the graph;
the processing subunit is used for performing matrix multiplication processing, linear transformation processing and nonlinear activation processing on the target incidence matrix and the object representation vector matrix to obtain an updated object representation vector matrix;
and the execution subunit is used for repeatedly executing preset times to perform matrix multiplication processing, linear transformation processing and nonlinear activation processing on the target incidence matrix and the updated object representation vector matrix to obtain an updated object representation vector matrix, so as to obtain the target object representation vector matrix.
In some embodiments, the propagation function of the graph neural network further comprises object characterization recall coefficients and propagation parameter recall coefficients, the processing subunit being configured to: and performing matrix multiplication processing and linear transformation processing on the target incidence matrix and the object characterization vector matrix, and supplementing the initial gradient information processing and nonlinear activation processing of the object characterization vector matrix according to the object characterization recall coefficient and the propagation parameter recall coefficient to obtain an updated object characterization vector matrix.
In some embodiments, the execution subunit is to: and repeatedly executing preset times to perform matrix multiplication processing and linear transformation processing on the target incidence matrix and the updated object characterization vector matrix, supplementing gradient information processing and nonlinear activation processing of the initial object characterization vector matrix according to the object characterization recall coefficient and the propagation parameter recall coefficient to obtain the updated object characterization vector matrix, and thus obtaining the target object characterization vector matrix.
The generating unit 303 is configured to generate label information corresponding to the target object characterization vector matrix.
In some embodiments, the constructing unit 303 is configured to: sequentially acquiring object representation vector information of each row in the target object representation vector matrix; tag information of the object characterization vector information of each line is generated.
And the training unit 304 is configured to train the preset neural network model through the target object characterization vector matrix and the corresponding label information, so as to obtain the trained preset neural network model.
In some embodiments, the training unit 304 is configured to: inputting the object characterization vector information of each line into the preset neural network model, and outputting corresponding prediction information; inputting the prediction information and the corresponding label information into a loss function; and carrying out iterative processing on the network parameters of the preset neural network model according to the loss values calculated by the loss function until convergence to obtain the trained preset neural network model.
And the detecting unit 305 is configured to perform identity detection on the target object to be recognized based on the trained preset neural network model.
In some embodiments, the detection unit 305 is configured to: acquiring a characteristic matrix to be recognized of an object set to be recognized and a target association matrix to be recognized formed by interactive behaviors; transmitting the information of the compactness between different objects to be detected in the set to be detected in the incidence matrix of the object to be recognized to the characteristic matrix to be recognized to obtain a characteristic vector matrix of the object to be recognized; and inputting the representation vector matrix of the target object to be recognized into the trained neural network model, and outputting the identity detection result of each target object to be recognized in the set of the target objects to be recognized.
The specific implementation of each unit can refer to the previous embodiment, and is not described herein again.
As can be seen from the above, in the embodiment of the present application, the obtaining unit 301 obtains the feature matrix of the set of objects to be detected and the target association matrix formed by the interaction behavior; the propagation unit 302 propagates the information representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the feature matrix to obtain a target object representation vector matrix; the generating unit 303 generates label information corresponding to the target object representation vector matrix; the training unit 304 trains the preset neural network model through the target object representation vector matrix and the corresponding label information to obtain the trained preset neural network model; the detection unit 305 performs identity detection on the target object to be recognized based on the trained preset neural network model. Therefore, intimacy information implicit in the target incidence matrix is transmitted to the feature matrix, label calibration and training are carried out, the trained preset neural network model can be combined with the neighbor object relation in the object set to carry out identity recognition, and the accuracy of information detection is greatly improved.
The embodiment of the present application further provides a computer device, as shown in fig. 5, which shows a schematic structural diagram of a server according to the embodiment of the present application, specifically:
the computer device may include components such as a processor 401 of one or more processing cores, memory 402 of one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will appreciate that the computer device configuration illustrated in FIG. 5 does not constitute a limitation of computer devices, and may include more or fewer components than those illustrated, or some components may be combined, or a different arrangement of components. Wherein:
the processor 401 is a control center of the computer device, connects various parts of the entire computer device using various interfaces and lines, and performs various functions of the computer device and processes data by running or executing software programs and/or modules stored in the memory 402 and calling data stored in the memory 402, thereby monitoring the computer device as a whole. Optionally, processor 401 may include one or more processing cores; optionally, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles operating systems, user interfaces, application programs, and the like, and the modem processor mainly handles wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 401.
The memory 402 may be used to store software programs and modules, and the processor 401 executes various functional applications and data processing by operating the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data created according to the use of the server, and the like. Further, the memory 402 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 access to the memory 402.
The computer device further comprises a power supply 403 for supplying power to the respective components, and optionally, the power supply 403 may be logically connected to the processor 401 through a power management system, so that functions of managing charging, discharging, power consumption, and the like are implemented through the power management system. The power supply 403 may also include any component of one or more dc or ac power sources, recharging systems, power failure detection circuitry, power converters or inverters, power status indicators, and the like.
The computer device may also include an input unit 404, which input unit 404 may be used to receive input numeric or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.
Although not shown, the computer device may further include a display unit and the like, which are not described in detail herein. Specifically, in this embodiment, the processor 401 in the computer device loads the executable file corresponding to the process of one or more application programs into the memory 402 according to the following instructions, and the processor 401 runs the application program stored in the memory 402, so as to implement the various method steps provided by the foregoing embodiments, as follows:
acquiring a characteristic matrix of an object set to be detected and a target incidence matrix formed by interactive behaviors; transmitting the information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the characteristic matrix to obtain a target object representation vector matrix; generating label information corresponding to the target object representation vector matrix; training a preset neural network model through the target object representation vector matrix and corresponding label information to obtain the trained preset neural network model; and carrying out identity detection on the target object to be recognized based on the trained preset neural network model.
In the above embodiments, the descriptions of the embodiments have respective emphasis, and parts that are not described in detail in a certain embodiment may refer to the above detailed description of the information detection method, and are not described herein again.
As can be seen from the above, the computer device in the embodiment of the present application may obtain a target association matrix formed by the feature matrix and the interaction behavior of the set of objects to be detected; transmitting information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the feature matrix to obtain a target object representation vector matrix; generating label information corresponding to the target object representation vector matrix; training the preset neural network model through the target object representation vector matrix and the corresponding label information to obtain the trained preset neural network model; and carrying out identity detection on the target object to be recognized based on the trained preset neural network model. Therefore, intimacy information implicit in the target incidence matrix is transmitted to the feature matrix, label calibration and training are carried out, the trained preset neural network model can be combined with the neighbor object relation in the object set to carry out identity recognition, and the accuracy of information detection is greatly improved.
It will be understood by those skilled in the art that all or part of the steps of the methods of the above embodiments may be performed by instructions or by associated hardware controlled by the instructions, which may be stored in a computer readable storage medium and loaded and executed by a processor.
To this end, embodiments of the present application provide a computer-readable storage medium, in which a plurality of instructions are stored, where the instructions can be loaded by a processor to execute the steps in any one of the information detection methods provided in the embodiments of the present application. For example, the instructions may perform the steps of:
acquiring a characteristic matrix of an object set to be detected and a target incidence matrix formed by interactive behaviors; transmitting the information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the characteristic matrix to obtain a target object representation vector matrix; generating label information corresponding to the target object representation vector matrix; training a preset neural network model through the target object representation vector matrix and corresponding label information to obtain the trained preset neural network model; and carrying out identity detection on the target object to be recognized based on the trained preset neural network model.
According to an aspect of the application, a computer program product or computer program is provided, comprising computer instructions, the computer instructions being stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to cause the computer device to perform the method provided in the various alternative implementations provided by the embodiments described above.
The above operations can be implemented in the foregoing embodiments, and are not described in detail herein.
Wherein the computer-readable storage medium may include: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.
Since the instructions stored in the computer-readable storage medium can execute the steps in any information detection method provided in the embodiments of the present application, the beneficial effects that can be achieved by any information detection method provided in the embodiments of the present application can be achieved, which are detailed in the foregoing embodiments and will not be described again here.
The above detailed description is provided for an information detection method, an information detection apparatus, and a computer-readable storage medium, which are provided by the embodiments of the present application, and specific examples are applied herein to explain the principles and implementations of the present application, and the descriptions of the above embodiments are only used to help understand the method and the core ideas of the present application; meanwhile, for those skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Claims (15)
1. An information detection method, comprising:
acquiring a characteristic matrix of an object set to be detected and a target incidence matrix formed by interactive behaviors;
transmitting the information for representing the compactness between different objects to be detected in the set to be detected in the target incidence matrix to the feature matrix to obtain a target object representation vector matrix;
generating label information corresponding to the target object representation vector matrix;
training a preset neural network model through the target object representation vector matrix and corresponding label information to obtain the trained preset neural network model;
and carrying out identity detection on the target object to be recognized based on the trained preset neural network model.
2. The information detection method according to claim 1, wherein the step of transmitting the information indicating the closeness between different objects to be detected in the set to be detected in the target association matrix to the feature matrix to obtain a target object representation vector matrix comprises:
determining the feature matrix as an object characterization vector matrix;
inputting the target incidence matrix and the object characterization vector matrix into a propagation function of a neural network of a graph;
matrix multiplication processing, linear transformation processing and nonlinear activation processing are carried out on the target incidence matrix and the object characterization vector matrix to obtain an updated object characterization vector matrix;
and repeatedly executing preset times to perform matrix multiplication processing, linear transformation processing and nonlinear activation processing on the target incidence matrix and the updated object characterization vector matrix to obtain an updated object characterization vector matrix, thereby obtaining the target object characterization vector matrix.
3. The information detection method according to claim 2, wherein the propagation function of the graph neural network further includes object characterization recall coefficients and propagation parameter recall coefficients, and the step of performing matrix multiplication processing, linear transformation processing, and nonlinear activation processing on the target association matrix and the object characterization vector matrix to obtain an updated object characterization vector matrix includes:
and performing matrix multiplication processing and linear transformation processing on the target incidence matrix and the object characterization vector matrix, and supplementing the initial gradient information processing and nonlinear activation processing of the object characterization vector matrix according to the object characterization recall coefficient and the propagation parameter recall coefficient to obtain an updated object characterization vector matrix.
4. The information detection method according to claim 3, wherein the step of repeatedly performing matrix multiplication processing, linear transformation processing, and nonlinear activation processing on the target correlation matrix and the updated object characterization vector matrix by a preset number of times to obtain the updated object characterization vector matrix, and the step of obtaining the target object characterization vector matrix comprises:
and repeatedly executing preset times to perform matrix multiplication processing and linear transformation processing on the target incidence matrix and the updated object characterization vector matrix, supplementing gradient information processing and nonlinear activation processing of the initial object characterization vector matrix according to the object characterization recall coefficient and the propagation parameter recall coefficient to obtain an updated object characterization vector matrix, and thus obtaining the target object characterization vector matrix.
5. The information detection method according to claim 1, wherein the step of obtaining the feature matrix of the set of objects to be detected and the target association matrix formed by the interactive behavior comprises:
acquiring a characteristic matrix of an object set to be detected;
acquiring a plurality of incidence matrixes formed by the interactive behaviors of an object set to be detected under different feature dimensions;
performing matrix standardization on each incidence matrix to obtain a plurality of incidence matrices subjected to matrix standardization;
and carrying out weighted aggregation on the plurality of incidence matrixes subjected to matrix standardization to obtain the target incidence matrix.
6. The information detection method according to any one of claims 1 to 5, wherein the step of generating label information corresponding to the target object characterization vector matrix includes:
sequentially acquiring object representation vector information of each row in the target object representation vector matrix;
tag information of the object characterization vector information of each line is generated.
7. The information detection method according to claim 6, wherein the step of training a preset neural network model through the target object characterization vector matrix and corresponding label information to obtain the trained preset neural network model comprises:
inputting the object characterization vector information of each line into the preset neural network model, and outputting corresponding prediction information;
inputting the prediction information and the corresponding label information into a loss function;
and carrying out iterative processing on the network parameters of the preset neural network model according to the loss values calculated by the loss function until convergence, so as to obtain the trained preset neural network model.
8. The information detection method according to any one of claims 1 to 5, wherein the step of performing identity detection on the target object to be recognized based on the trained preset neural network model includes:
acquiring a characteristic matrix to be recognized of an object set to be recognized and a target association matrix to be recognized formed by interactive behaviors;
transmitting the information of the compactness between different objects to be detected in the set to be detected in the incidence matrix of the object to be recognized to the characteristic matrix to be recognized to obtain a characteristic vector matrix of the object to be recognized;
and inputting the representation vector matrix of the target object to be recognized into the trained neural network model, and outputting the identity detection result of each target object to be recognized in the set of the target objects to be recognized.
9. An information detecting apparatus, characterized by comprising:
the acquisition unit is used for acquiring a characteristic matrix of the set of objects to be detected and a target incidence matrix formed by interactive behaviors;
the propagation unit is used for propagating the compactness information which characterizes different objects to be detected in the set to be detected in the target incidence matrix to the characteristic matrix to obtain a target object characterization vector matrix;
the generating unit is used for generating label information corresponding to the target object representation vector matrix;
the training unit is used for training a preset neural network model through the target object representation vector matrix and the corresponding label information to obtain the trained preset neural network model;
and the detection unit is used for carrying out identity detection on the target object to be recognized based on the trained preset neural network model.
10. The information detection apparatus according to claim 9, wherein the propagation unit includes:
the determining subunit is used for determining the feature matrix as an object characterization vector matrix;
the input subunit is used for inputting the target incidence matrix and the object characterization vector matrix into a propagation function of the neural network of the graph;
the processing subunit is used for performing matrix multiplication processing, linear transformation processing and nonlinear activation processing on the target incidence matrix and the object representation vector matrix to obtain an updated object representation vector matrix;
and the execution subunit is used for repeatedly executing preset times to perform matrix multiplication processing, linear transformation processing and nonlinear activation processing on the target incidence matrix and the updated object representation vector matrix to obtain an updated object representation vector matrix, so as to obtain the target object representation vector matrix.
11. The information detection apparatus according to claim 10, wherein the propagation function of the graph neural network further includes object characterization recall coefficients and propagation parameter recall coefficients, and the processing subunit is configured to:
and performing matrix multiplication processing and linear transformation processing on the target incidence matrix and the object characterization vector matrix, and supplementing the initial gradient information processing and nonlinear activation processing of the object characterization vector matrix according to the object characterization recall coefficient and the propagation parameter recall coefficient to obtain an updated object characterization vector matrix.
12. The information detecting apparatus according to claim 11, wherein the executing subunit is configured to:
and repeatedly executing preset times to perform matrix multiplication processing and linear transformation processing on the target incidence matrix and the updated object characterization vector matrix, supplementing gradient information processing and nonlinear activation processing of the initial object characterization vector matrix according to the object characterization recall coefficient and the propagation parameter recall coefficient to obtain an updated object characterization vector matrix, and thus obtaining the target object characterization vector matrix.
13. The information detection apparatus according to claim 9, wherein the acquisition unit is configured to:
acquiring a characteristic matrix of an object set to be detected;
acquiring a plurality of incidence matrixes formed by the interactive behaviors of an object set to be detected under different feature dimensions;
performing matrix standardization on each incidence matrix to obtain a plurality of incidence matrices subjected to matrix standardization;
and carrying out weighted aggregation on the plurality of incidence matrixes subjected to matrix standardization to obtain the target incidence matrix.
14. The information detection apparatus according to any one of claims 9 to 13, wherein the generation unit is configured to:
sequentially acquiring object representation vector information of each row in the target object representation vector matrix;
tag information of the object characterization vector information of each line is generated.
15. A computer-readable storage medium storing a plurality of instructions adapted to be loaded by a processor to perform the steps of the information detection method according to any one of claims 1 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010884149.6A CN112044082B (en) | 2020-08-28 | 2020-08-28 | Information detection method and device and computer readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010884149.6A CN112044082B (en) | 2020-08-28 | 2020-08-28 | Information detection method and device and computer readable storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112044082A true CN112044082A (en) | 2020-12-08 |
CN112044082B CN112044082B (en) | 2021-11-30 |
Family
ID=73606461
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010884149.6A Active CN112044082B (en) | 2020-08-28 | 2020-08-28 | Information detection method and device and computer readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112044082B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113159778A (en) * | 2020-12-24 | 2021-07-23 | 西安四叶草信息技术有限公司 | Financial fraud detection method and device |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10112113B2 (en) * | 2016-03-30 | 2018-10-30 | Sony Interactive Entertainment Inc. | Personalized data driven game training system |
CN108769096A (en) * | 2018-03-30 | 2018-11-06 | 广东小天才科技有限公司 | Remote monitoring method, device, equipment and storage medium for mobile equipment |
CN109102007A (en) * | 2018-07-26 | 2018-12-28 | 深圳市腾讯网络信息技术有限公司 | Exchange method, device, terminal and storage medium based on label |
CN109214444A (en) * | 2018-08-24 | 2019-01-15 | 小沃科技有限公司 | Game Anti-addiction decision-making system and method based on twin neural network and GMM |
CN109771952A (en) * | 2018-12-28 | 2019-05-21 | 努比亚技术有限公司 | Based reminding method, terminal and computer readable storage medium based on game fatigue strength |
CN110163611A (en) * | 2019-03-18 | 2019-08-23 | 腾讯科技(深圳)有限公司 | A kind of personal identification method, device and relevant device |
CN111582538A (en) * | 2020-03-25 | 2020-08-25 | 清华大学 | Community value prediction method and system based on graph neural network |
-
2020
- 2020-08-28 CN CN202010884149.6A patent/CN112044082B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10112113B2 (en) * | 2016-03-30 | 2018-10-30 | Sony Interactive Entertainment Inc. | Personalized data driven game training system |
CN108769096A (en) * | 2018-03-30 | 2018-11-06 | 广东小天才科技有限公司 | Remote monitoring method, device, equipment and storage medium for mobile equipment |
CN109102007A (en) * | 2018-07-26 | 2018-12-28 | 深圳市腾讯网络信息技术有限公司 | Exchange method, device, terminal and storage medium based on label |
CN109214444A (en) * | 2018-08-24 | 2019-01-15 | 小沃科技有限公司 | Game Anti-addiction decision-making system and method based on twin neural network and GMM |
CN109771952A (en) * | 2018-12-28 | 2019-05-21 | 努比亚技术有限公司 | Based reminding method, terminal and computer readable storage medium based on game fatigue strength |
CN110163611A (en) * | 2019-03-18 | 2019-08-23 | 腾讯科技(深圳)有限公司 | A kind of personal identification method, device and relevant device |
CN111582538A (en) * | 2020-03-25 | 2020-08-25 | 清华大学 | Community value prediction method and system based on graph neural network |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113159778A (en) * | 2020-12-24 | 2021-07-23 | 西安四叶草信息技术有限公司 | Financial fraud detection method and device |
CN113159778B (en) * | 2020-12-24 | 2023-11-24 | 西安四叶草信息技术有限公司 | Financial fraud detection method and device |
Also Published As
Publication number | Publication date |
---|---|
CN112044082B (en) | 2021-11-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113408743B (en) | Method and device for generating federal model, electronic equipment and storage medium | |
CN111632379B (en) | Game role behavior control method and device, storage medium and electronic equipment | |
CN112221159B (en) | Virtual item recommendation method and device and computer readable storage medium | |
CN111282267A (en) | Information processing method, information processing apparatus, information processing medium, and electronic device | |
CN111666919A (en) | Object identification method and device, computer equipment and storage medium | |
CN112395515B (en) | Information recommendation method and device, computer equipment and storage medium | |
CN112733043B (en) | Comment recommendation method and device | |
CN112434791A (en) | Multi-agent strong countermeasure simulation method and device and electronic equipment | |
CN111310918A (en) | Data processing method and device, computer equipment and storage medium | |
CN111340112A (en) | Classification method, classification device and server | |
CN112044082B (en) | Information detection method and device and computer readable storage medium | |
CN115345301A (en) | Model training method and device based on mixed disturbance, storage medium and server | |
CN117218482A (en) | Model training method, video processing device and electronic equipment | |
CN115168720A (en) | Content interaction prediction method and related equipment | |
CN113230650B (en) | Data processing method and device and computer readable storage medium | |
CN115994321A (en) | Object classification method and related device | |
CN112274935A (en) | AI model training method, use method, computer device and storage medium | |
CN116522131A (en) | Object representation method, device, electronic equipment and computer readable storage medium | |
CN111935259B (en) | Method and device for determining target account set, storage medium and electronic equipment | |
CN113010772B (en) | Data processing method, related equipment and computer readable storage medium | |
CN115168722A (en) | Content interaction prediction method and related equipment | |
CN114119997A (en) | Training method and device for image feature extraction model, server and storage medium | |
CN113476833A (en) | Game action recognition method and device, electronic equipment and storage medium | |
CN118051782B (en) | Model training method, business processing method and related device | |
CN116933837A (en) | Data processing method, device, electronic equipment, program product and medium |
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 | ||
REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40035463 Country of ref document: HK |
|
GR01 | Patent grant | ||
GR01 | Patent grant |