159 results about "Learning abilities" patented technology

The invention relates to the technical field of artificial intelligence, can be applied to business scenes of medical health, financial science and technology and the like, and discloses a voice generation method which comprises the following steps: constructing a multi-language voice synthesis model, obtaining plain text data and paired voice text data, and constructing an expansion vocabulary; updating a language perception embedding layer and model parameters, and converting an input text into a mark sequence; and the encoder extracts context semantic features, extracts pronunciation rule features, and the decoder fuses the features to generate an acoustic feature sequence, and converts the acoustic feature sequence into target voice data. According to the invention, the multi-language speech synthesis model is combined with the language perception embedding layer, so that the speech generation capability of a low-resource language is improved; the text conversion accuracy is improved by expanding the vocabulary, the target language learning ability is enhanced by unsupervised training, the low data environment adaptability is optimized by supervised training, and the speech naturalness and fluency are improved by feature fusion.

The invention discloses an intelligent programming assistance method and system combining a large model and a knowledge graph. The method comprises the steps that the system is initialized, and a programming domain knowledge graph body and an education programming large language model are loaded; constructing a learner initial portrait and associating the learner initial portrait with a knowledge graph; personalized programming learning experience is provided, including real-time programming tutoring, and thinking chain-based heuristic guidance is generated through combination of a large language model and knowledge graph information; and interactive feedback and system iteration are carried out, and a reinforcement learning mechanism based on human feedback is utilized to continuously optimize the large language model. Through deep fusion of structured knowledge of a knowledge graph and understanding generation ability of a large language model, the method aims at solving the problems that an existing programming auxiliary tool is insufficient in individuation and heuristic, heterogeneous knowledge is difficult to fuse, teaching resource modes are single, and an effective evolution mechanism is lacked. More accurate, personalized, heuristic and sustainable-evolution intelligent programming assistance can be provided, and the programming thinking and autonomous learning ability of learners can be effectively improved.

The invention relates to the technical field of educational resource recommendation, and discloses an educational resource intelligent recommendation method and system based on big data driving, and the system comprises a data fusion processing module, a portrait modeling module, a resource feature engine module, an intelligent recommendation core module and a closed-loop feedback module. By fusing knowledge state, cognitive ability and interest preference three-dimensional portraits, capturing user learning ability evolution in real time, dynamically updating knowledge mastery by adopting a knowledge tracking model, and perceiving interest migration in combination with an attention mechanism, the problem of learning cold start in a new field is solved, recommendation coverage range and accuracy are improved, and user experience is improved. The cognitive load sensitive ant colony optimization algorithm is designed, the learning path continuity is guaranteed through a heuristic function, the learning path structure reasonability is optimized, the cognitive burden of a user is reduced, a personalized knowledge attenuation model is constructed by fusing an Ebbinghaus forgetting curve, the knowledge long-term retention rate is increased, and the review resource release accuracy is enhanced.

The invention provides a personalized learning path planning system and method, and belongs to the technical field of emerging software and emerging technical services, and the system comprises a data acquisition module which is used for collecting learning behavior data of a terminal user, obtaining a learner portrait package based on the learning behavior data, and sending the learner portrait package to a server; the path planning module is used for carrying out node matching on the learning ability feature vector and a pre-constructed knowledge and skill map, outputting a to-be-learned content node set, dividing learning advanced levels and generating learning path description containing to-be-learned nodes and the advanced levels, and the path generation module is used for generating a primary path planning scheme, the scheme feasibility verification module is used for carrying out scheme feasibility verification in combination with the learning advanced level and the path adaptation parameters, and then outputting a final executable path scheme, and the scheme execution module is used for executing the final executable path scheme and controlling learning content pushing and progress adjustment. The problems that in the prior art, personalized learning path planning is not high in precision, not high in adaptability, lack of dynamic optimization and the like are solved.

The invention relates to the technical field of education, in particular to an English personalized learning recommendation method based on big data, which comprises the following steps: collecting original data of English homework completion speed and answer accuracy of a learner, calculating average completion time and accuracy by using a statistical analysis method, identifying the deviation between learning ability and interest points, and recommending the learning ability to the learner. And obtaining a learning ability evaluation result. According to the invention, through predicting learning content demands, not only is the progress of a learner captured, but also future demands can be predicted, so that education resources are prepared in advance, prospective matching of teaching contents is realized, and through optimizing a teaching material sequence and contents, high matching of teaching materials and personalized learning demands is ensured; the use efficiency of educational resources and the maximization of the learning effect are remarkably improved, the learning path is dynamically adjusted, the learning adaptability is evaluated, the learning process is optimized, the flexible application of course content is enhanced, and more personalized learning experience and higher learning achievement are achieved.

The invention discloses a reward model training method, a large language model optimization method and correlation, and the reward model training method comprises the steps: obtaining a preference training sample pair and a to-be-trained reward model, the preference training sample pair comprising a preferred response sample and a non-preferred response sample; calculating an award score difference between the preferred response sample and the non-preferred response sample based on a to-be-trained award model; constructing a cost matrix based on the reward score difference and the semantic association degree between the preferred response sample and the non-preferred response sample; calculating a loss margin based on the cost matrix; and based on the loss margins, carrying out calculation to obtain paired preference loss values of the band margins, and updating parameters of the to-be-trained reward model by taking minimization of the loss values based on the band margins as an optimization target to obtain a trained reward model. The learning ability and overall generalization performance of the model for difficult samples are improved, excessive dependence on simple samples is avoided, and then the generation quality of the large language model in complex tasks is improved.

The invention provides a multi-element ability evaluation system and evaluation method for children education, and relates to the field of children education systems. According to the multi-element ability evaluation system for children education, the multi-element ability evaluation system for children education is realized through a multi-element ability evaluation system, and the multi-element ability evaluation system comprises three functional modules, namely an evaluation module, a management module and an evaluation analysis module, the evaluation module comprises a learning ability evaluation module, a physical health evaluation module and an emotion and mental state evaluation module, the evaluation module, the evaluation analysis module and the management module are in communication connection through the Internet, and the multi-element ability evaluation system comprises an evaluation report generation module. According to the invention, comprehensive test evaluation can be carried out on three aspects of intelligence quotient and physical ability of the children, so that multiple abilities of the children can be known more comprehensively through the evaluation result, the children can be better taught according to the case and the materials, and the education of the children is more efficient.

The invention provides a learning performance precise teaching management method based on time sequence behavior modeling. The learning performance precise teaching management method comprises the following steps: S1, carrying out multi-modal learning behavior data acquisition; s2, aligning and segmenting the time sequence data; s3, performing knowledge retention rate differential modeling; s4, carrying out time sequence feature extraction; s5, constructing a mixed time sequence model; s6, performing dynamic learning ability evaluation; s7, carrying out adaptive resource recommendation; s8, carrying out cross-correction knowledge diffusion optimization; the method has the following advantages: the method is real-time and accurate; the data is acquired to acquire the intervened closed-loop delay lt; compared with the traditional method, the time is increased by 40 times. And cost optimization: the workload of teachers is reduced by 58% and the resource purchase cost is reduced by 42% through an automation strategy. The scale effect is that federal learning supports ten-thousand-person-level concurrence, and the model updating period is shortened to the hour level from the quarter level. Education fairness: cross-school knowledge diffusion enables the superior rate of weak schools to be improved by 29%.

The invention discloses a multi-dimensional capability construction knowledge tracking method based on learning behavior decoupling, and belongs to the technical field of knowledge tracking. The method comprises the steps of obtaining test question embedding and test question-response embedding, and obtaining learning deviation behavior characterization and multi-scale behavior characterization based on the test question-response embedding; inputting the learning deviation behavior representation and the multi-scale behavior representation into a binary state model to obtain a knowledge construction representation; obtaining a test question sequence structure representation based on test question embedding, and performing multi-scale association on the test question sequence structure representation to obtain a multi-scale test question association representation; the multi-scale test question association representation and the knowledge construction representation are subjected to student ability modeling through frequency domain perception fusion, and differential learning ability is obtained; and predicting the student test question performance based on the differentiated learning ability to obtain a student test question performance prediction value. The method improves the accuracy of student test question performance prediction.

The invention relates to the technical field of data processing, in particular to a hyper-relation extraction method based on collaboration of multiple large language models and application of the hyper-relation extraction method. According to the method, the extraction cue word serving as the extraction rule is optimized by means of the strong learning ability of the large language model, a correction mechanism is given to the extraction cue word to correct the generated hyper-relationship, and rapid and accurate text hyper-relationship extraction can be realized without performing fine adjustment on parameters of the model through supervised learning. The method aims at solving the problem of how to improve the super-relation extraction efficiency of the text.

The invention discloses a collaborative training method, medium and device based on an LLM large and small model, and the method comprises the steps: S1, obtaining a huge knowledge system through the combination of the language understanding capability and pre-training capability of a general large model, and carrying out the cold start of an AI service 0 sample needed by a business scene, and enabling the AI service 0 sample to be online; s2, performing a small amount of annotation on sample data generated by a business scene, performing fine tuning on the general large model to form a scene large model, performing continuous learning to enable the scene large model to have knowledge in the field, and quickly improving an algorithm effect; and S3, distilling knowledge in the field obtained by the scene large model into a plurality of small models, and fusing results of the plurality of small models by using a scoring mechanism to realize collaborative training of the large and small models. According to the method, any scene text service can be subjected to cold start online under the condition of limited hardware resources, large and small model cooperative training of large model knowledge can be obtained through a small amount of labeling, and the model learning ability and the working efficiency are greatly improved.

The invention discloses an automatic game testing method and system, and the method achieves the automatic generation and execution of testing logic through the building of a knowledge base and an event base, and the cooperative management of the knowledge base and the event base, and reduces the manual intervention. Converting a natural language demand input by a user side into a semantic feature, and matching the semantic feature with an operation instruction in an event library to obtain an executable test logic so as to reduce a test threshold; triggering the event in the event library according to the executable test logic, pushing the execution result of the event to the user side, and updating the knowledge base and the event library by using the execution feedback data of the event, thereby forming an execution-feedback-update closed loop, and improving the self-learning ability of the test system. In this way, the logic rules in the game development kit are converted into the knowledge base, intelligent analysis and dynamic feedback optimization of natural language requirements are achieved in combination with the preset operation instruction of the event base, and the test efficiency and the adaptive capacity are remarkably improved.

The invention discloses a teacher teaching ability improvement training method and system, and the method comprises the steps: determining a plurality of evaluation dimensions for teacher teaching ability, obtaining a quantitative evaluation index of each evaluation dimension, and obtaining a plurality of high-quality training course resources according to the quantitative evaluation indexes; historical teaching state parameters input by each teacher are obtained, and teaching ability evaluation is carried out on each teacher; analyzing a teaching short board index and short board state description of each teacher according to an evaluation result, and determining capability improvement demand parameters; and screening out a target training course resource from the plurality of high-quality training course resources according to the ability improvement demand parameters, making a training plan based on the learning ability information of each teacher, and carrying out teaching ability improvement training on each teacher according to the training plan. According to the invention, accurate short board related ability training can be accurately carried out for teachers with different teaching levels, the training efficiency and stability are improved, the adaptability of training resources is ensured, and the practicability is improved.

The invention discloses a method, a system, equipment and a medium for automatically updating cue words of a large language model system, belongs to the technical field of cue word optimization of large language models, and aims to solve the technical problem of how to optimize cue words of large language models on the basis of recent interaction historical sessions between users and large language models or agents. Automatic updating of cue words of a large language model system is achieved, the continuous and lifelong learning ability of the large language model is improved, and then the thinking depth of the large language model is improved. According to the technical scheme, the method comprises the steps of obtaining a system cue word updating model with an interactive historical text compression memory function through SFT supervised fine tuning or enhanced post-training based on a system cue data set, constructing a system cue word cue experience entry knowledge base based on the system cue word updating model and in combination with long-time interactive dialogue history of a user and a large language model, and obtaining the system cue word cue experience entry knowledge base. And on the basis of a system cue word updating model, the current dialogue history is interacted by combining the user and the large language model, so that the system cue word is automatically updated.

The invention relates to the technical field of heart failure auxiliary diagnosis and treatment, in particular to a heart failure auxiliary diagnosis and treatment knowledge distillation method and system.The knowledge distillation method comprises the steps that an age-crossing fusion data set is built by combining ancient medical case text data related to traditional Chinese medicine heart failure; performing field fine tuning on the pre-training model to generate a basic teacher model; performing semantic alignment, and outputting a cross-time contrast database; constructing a curative effect scoring model by using the time sequence model, and outputting a quantitative association database; using a graph neural network to output an intelligent heart failure knowledge graph, and generating a comprehensive teacher model; performing knowledge distillation on the comprehensive teacher model to generate a clinical student model; the knowledge distillation system is applied to the knowledge distillation method, has a dynamic learning ability, can fuse ancient and modern trans-time knowledge and combine traditional Chinese medicine ancient book wisdom to carry out heart failure auxiliary diagnosis and treatment, and can give consideration to model performance and deployment efficiency through knowledge distillation.

The present invention relates to the technical field of knowledge tracing, and in particular to a method and system for predicting learning conditions of learners based on meta-paths, and an electronic device. In the method, after learner data, exercise book data, and learner answer interaction sequence data are obtained, firstly, a learner behavioral graph is constructed, and learning content preferences of learners are extracted; secondly, learning abilities of the learners are calculated; thirdly, features of the learning content preferences and features of the learning abilities are fused, and a learner heterogeneous information network based on meta-paths is constructed; next, embedding learning is performed on the leaner heterogeneous information network by using a layered graph attention mechanism, to extract leaner fused features; and finally, learning conditions of the learners are predicted by using a GRU-based knowledge tracing model. Compared with the prior art, the present invention has the advantages of more accurately predicting the probability that a learner will answer a next exercise correctly, evaluating and predicting a learning outcome of the learner, and thus providing the learner with multi-dimensional feedback and guidance.

The invention relates to the technical field of natural language processing, in particular to a data enhancement and feature enhancement-based Chinese named entity recognition method, which is based on a feature enhancement double-attention named entity recognition model, and is characterized in that the model comprises an embedded layer, a data enhancement module, a multi-scale convolution fusion attention layer and a prediction layer; the embedded layer uses a dual-channel attention fusion module to process texts in parallel, fuses multi-dimensional information features of Chinese characters, combines an error correction type mask language model, a pre-training model and a bidirectional gating loop unit, fuses local and global text features, and obtains text representation from multiple dimensions and multiple levels; according to the method, by introducing an innovative model mechanism or training strategy, challenges such as label sparsity and text noise existing in a Chinese named entity recognition task can be effectively handled, the learning ability and generalization performance of the model for long-tail entity categories are improved, and the robustness of the model in a real and non-ideal data environment is enhanced.

The invention discloses an efficient simultaneous interpretation method based on an expert routing threshold, and relates to the field of voice processing, the method is based on a classical Transform architecture model, an expert routing strategy model based on the expert routing threshold is constructed, and multi-language streaming translation is realized, the model comprises a streaming voice encoder, and the streaming voice encoder adopts a hybrid design and is connected with the expert routing threshold. The block-by-block autoregression block is composed of an autoregression block and a non-autoregression block; the text decoder is used for simultaneously processing the complete offline voice and the randomly truncated voice prefix to generate a hidden state; the routing threshold module is realized by a feedforward network and projects the final hidden state into a scalar value to determine an expert weight; and the hybrid expert post-processing module shares a language model head with the text decoder, and predicts a target translation sequence in combination with prefix information and global information. According to the method, a mixed expert threshold scheme is adopted to learn the strategy, the self-learning ability of the neural network is fully played, good effects are achieved in streaming translation and streaming TTS, and the method can be used for generating more streaming sequences.

The invention discloses an intelligent learning asking method and system based on a large language model, and relates to the technical field of generative artificial intelligence and intelligent education, and the method comprises the steps: guiding a user to carry out deep interaction through a human-intelligent interaction module, so as to generate a question and answer set containing multi-party intelligence; a teacher user purifies a question and answer set through a question and answer filing unit to obtain high-quality data, and the system automatically extracts knowledge from dialogues by using the complete interaction process to dynamically update a multilevel knowledge base composed of a personalized knowledge graph and a global knowledge graph; meanwhile, model fine adjustment is conducted on an AI intelligent agent based on a cognitive framework and a reasoning framework through archived high-quality data, and a closed loop of double evolution of a knowledge base and model parameters is formed. According to the invention, the professional ability of AI in the vertical field can be significantly improved, and the high-order learning ability of students can be effectively cultivated.

The invention belongs to the field of resource recommendation, and particularly relates to a learning resource recommendation method, system and device and a storage medium, the learning time of a user is divided into a fragmented time range and a long-time learning time range, the learning ability of the user is calculated, and according to the learning ability and the association degree of each chapter in a knowledge graph, the learning resource is recommended. When the current time is in the fragmented learning time range, a recommendation emergency value is obtained according to the current learning urgency degree and the knowledge gap risk degree, when the recommendation emergency value is smaller than a preset threshold value, a chapter with a maximum recommendation strategy is matched in the knowledge graph for recommendation, and if the recommendation emergency value is larger than the preset threshold value, the recommendation strategy is matched with the chapter with the maximum recommendation strategy. And if yes, optimizing and recommending chapters corresponding to the urgently needed learning resource type for learning. According to the method and the device, the corresponding learning resource recommendation is carried out in consideration of the time fragments of user learning, and the content which is urgently needed to be learned at present is also considered during recommendation to carry out preferential learning, so that the examination score is improved.

The invention provides a method and system for cognitive diagnosis and attribution analysis of teacher teaching, and belongs to the technical field of teaching management. The method specifically comprises the following steps: constructing a hybrid multi-dimensional capability vector for a student agent; acquiring a teaching instruction of a teacher, analyzing the teaching instruction, determining a teaching mode and analyzing teaching content; the intelligent agent responds to the teaching mode and the teaching content, calls an ability probe algorithm to evaluate the performance of the intelligent agent for completing the task, and calculates and updates a multi-dimensional ability vector of the intelligent agent; and comparing state differences of the multi-dimensional capability vector before and after evolution, modifying a teaching instruction based on an anti-fact attribution algorithm, carrying out anti-fact inference, and generating an evaluation report containing diagnosis information. By constructing a hybrid multi-dimensional ability vector, accurate description and dynamic tracking of the learning ability of the student are realized; and evaluating task performance in real time and updating a multi-dimensional capability vector through a capability probe algorithm to form a closed-loop feedback mechanism.

The invention discloses a reinforcement learning intrusion detection method based on space-time attention and dynamic courses, and the method comprises the steps: carrying out the dynamic weight distribution of input features in space and time dimensions through a multi-level space-time attention mechanism, and enhancing the expression capability of key features; a dynamic curriculum learning strategy is adopted to realize progressive difficulty adjustment of training samples, and the learning ability of the model is gradually improved through initial difficulty estimation, dynamic threshold adjustment and curriculum sample selection; designing a composite reward mechanism to optimize a reward signal, and combining a basic classification reward and a stability reward to promote stable convergence of the strategy; a self-evolution target network module is introduced, and dynamic update management of a target network is realized through performance monitoring, emergency update triggering and frequency self-adaption. A space-time cooperative detection closed loop is constructed, and an end-to-end intrusion detection process is realized through environment interaction, strategy optimization and online detection. The method effectively improves the accuracy and robustness of intrusion detection, and adapts to a complex network attack scene.

The invention relates to the technical field of learning activity optimization, and discloses an interactive learning activity dynamic optimization method and system, and the method comprises the steps: obtaining learning data of a user, extracting an interest label and a learning capability label of the user through a preset user behavior analysis model, and generating a user state label; generating a corresponding activity generation path based on the user state tag in combination with a preset course knowledge base; matching a corresponding learning activity according to the activity generation path to obtain an initial learning activity; optimizing the activity generation path by adding a path branch, trimming the path branch or replacing a path node in combination with a real-time learning state of the user to obtain an activity optimization path; optimizing the initial learning activity through the activity optimization path to obtain an optimized learning activity so as to dynamically optimize the interactive learning activity; according to the method, the learning activity can be dynamically optimized, and the state change condition of the user in the learning process can be responded in real time.

The invention relates to an RDFS ontology-oriented large language model evaluation and cue word optimization method. The method comprises the following steps of 1, constructing a positive example data set and a negative example data set; 2, designing cue word templates of multiple levels; 3, test cases in the standardized test data set are combined with different types of cue word templates, an input sequence is generated and sent to a to-be-evaluated large language model, and an output judgment result and an explanation text are obtained; 4, analyzing the judgment result and the explanation text to generate a comprehensive evaluation report; and 5, based on a data analysis result of the comprehensive evaluation report, forming a system performance map by constructing a'cue word type-evaluation dimension-axiom type 'ternary mapping relationship, and implementing an optimization process according to the system performance map. According to the scheme, the problems of incomplete evaluation system, imperfect test data, single evaluation dimension and the like in the prior art are solved, and systematic and precise evaluation and improvement of the model RDFS ontology learning ability are realized.

The invention discloses a learning path optimization method and system based on machine learning, and the method comprises the steps: collecting the learning data of a learner, and enabling the learning data to comprise a learning object and the information of the learner; establishing a double-layer knowledge graph according to the knowledge type of the learning object; collecting behavior information data of the learner, evaluating a learning style of the learner user according to the behavior information data of the learner, and analyzing and dynamically calculating a learning ability value of the learner and a difficulty value of learning resources in real time according to the learning behavior information data; according to the ability value of the learner, the difficulty value of the learning resource and the information of the learner, generating an optimal learning path by utilizing a double-layer knowledge graph, monitoring the current learning duration, score and participation degree change of the learner in real time, and circularly feeding back and updating the learning path, so that the learning efficiency is improved. And higher-quality and more personalized learning path recommendation is provided for learners.

The invention discloses a personalized art teaching capability evaluation and course generation method and system, and the system comprises a handwriting dynamics analysis module which is used for collecting handwriting dynamics data in the creation process of a learner, and generating handwriting feature data; the learner portrait construction module is used for receiving the handwriting feature data and constructing a five-dimensional learner portrait including art ability, learning ability, psychological features, professional interests and personal aesthetic appreciation in combination with the basic information of the learner; the knowledge graph module is used for constructing a knowledge graph in the art professional field; the ability evaluation module is used for generating an ability evaluation result of the learner based on the five-dimensional learner portrait and the knowledge graph by adopting a three-layer neural network structure; and the course generation module is used for classifying the learners into a single-subject type, a single-ability type, a single-project type or a comprehensive type according to ability evaluation results of the learners, and generating corresponding personalized learning paths and course contents based on classification types.

The invention discloses an intelligent foreign language teaching resource planning method and system based on a knowledge graph, and relates to the technical field of foreign language teaching and intelligence. The planning method comprises the following steps: constructing a knowledge system and data acquisition: collecting foreign language teaching knowledge, carrying out preprocessing, constructing a foreign language knowledge graph by means of a graph database, and carrying out data acquisition; the method comprises the following steps: establishing an association relationship between knowledge, acquiring learning feature data of learning basis, style, progress and preference of students through student registration information, learning behavior data acquisition and learning ability test, and acquiring daily curriculum schedule and idle time data of the students in real time; the foreign language knowledge graph is constructed, scattered teaching knowledge is systematically integrated, network association among knowledge is formed, and learning requirements of students can be accurately positioned and intelligent screening and efficient adaptation of teaching resources can be realized in combination with learning feature data, including learning basis, style, progress and preference, of the students.

The embodiment of the invention discloses a data processing method and device based on a large language model. According to the method, a to-be-scored question group is obtained, the to-be-scored question group comprises questions, answers of answering parties and reference answers, and according to the to-be-scored question group and key words, a prompt template, a false question generation prompt template and an evaluation generation prompt template are extracted, respectively constructing a keyword extraction prompt statement, a pseudo question generation prompt statement and an evaluation generation prompt statement, respectively inputting the constructed statements into a large language model to obtain corresponding keyword groups, question groups and global evaluation, determining text similarity of answers of answering parties and reference answers, and determining the similarity of the answers of the answering parties and the reference answers; and inputting the keyword group, the question group, the global evaluation and the text similarity into a pre-trained comprehensive scoring model to obtain a score corresponding to the answer of the answering party. According to the method, the advantages of the large language model and the comprehensive scoring model are combined, the generalization generation capability of the large language model is exerted, and the vertical domain scene learning capability of the comprehensive scoring model is utilized.

The invention provides a training method of a face recognition model, and a face recognition method and system, and relates to the technical field of face recognition, and the method comprises the steps: pre-training a teacher network; constructing a student network and establishing a knowledge distillation framework; dividing the feature vector output by the teacher network into blocks, and applying different weights according to the importance of each block so as to form weighted block mean square error loss; the cosine similarity of feature vectors output by the teacher network and the student network is used as a difficult sample indicator; the angle interval in additive angle interval loss is dynamically adjusted according to the indicator, and differential constraints are applied to samples with different difficulty levels. According to the method, the learning ability for key features is enhanced through weighted block mean square error loss, attention distribution for difficult and simple samples is optimized through dynamic additive angle interval loss, the identification precision, convergence speed and model compression effect of a student network are effectively improved, and the method is suitable for resource-limited edge device deployment.

The invention provides an adversarial learning-based data value protection method facing transverse federated learning, and aims to prevent a model from being illegally migrated to an unauthorized task for use. The method comprises the following steps: locally extracting intermediate features by a client, and inputting the intermediate features into a main task classifier and a plurality of adversarial task classifiers; a gradient inversion layer is introduced, so that the feature extractor enhances the identification capability of a main task in training, and weakens the generalization capability of an unauthorized task at the same time; a composite loss function is constructed, and optimization balance among tasks is realized through dynamic weight adjustment; and finally, carrying out model parameter aggregation and iterative updating under a federated learning framework. According to the method, on the premise that the performance of a main task is not affected, the learning ability of the model for sensitive attributes is effectively limited, so that unexpected leakage of data values is inhibited, and the safety and specificity of the model are improved.