342 results about "Knowledge transfer" patented technology

The invention discloses a distribution line unmanned aerial vehicle edge end real-time visual identification and target detection method and system based on lightweight AI. The method is based on a YOLOv8 architecture, and constructs a complete lightweight detection framework comprising a feature extractor, an enhancement module and a simplified detection head by introducing a space structure maintaining assembly, a separated large kernel convolution and a weighted reconstruction feature pyramid. A cross-dimension semantic relation model is constructed by utilizing a combined attention structure, multi-level feature fusion is realized by reconstructing a feature pyramid network, model training is performed by adopting a hybrid optimization function and a dynamic sample adjustment mechanism, and the model is deployed on a mobile computing platform after being optimized by a hierarchical knowledge transfer method. According to the method, the detection speed is remarkably increased while high precision is kept, and real-time identification and anomaly analysis of the power line element are effectively realized.

The invention discloses an energy storage battery health feature extraction and state evaluation method based on transfer learning. The method comprises the following steps: S1, constructing a source domain health feature library and pre-training a model; according to the method, dependence on complete cyclic data is broken through, high precision and robustness are still achieved under the conditions of data sparsity and working condition difference, and the method is suitable for intelligent operation and maintenance and predictive maintenance of an energy storage power station. And meanwhile, common incomplete and partial charge and discharge data fragments under actual working conditions can be directly utilized for feature extraction and state evaluation, dependence on complete charge and discharge cycles is avoided, and the application scene of the data driving method is greatly widened.

The invention discloses a wind power prediction method based on multi-source domain deep transfer learning, and relates to the field of new energy power prediction.The method comprises the steps that the data distribution difference between a multi-source domain and a target domain is reduced through an Euclidean alignment method, and the maximum mean value difference between the domains after optimization is obtained; setting a migration weight factor for each source domain based on the maximum mean value difference, and constructing a weighted migration training set; constructing a multi-source domain deep migration learning model, and carrying out migration training on the model based on a weighted migration training set; and finely tuning the pre-training model according to a small amount of data of the target domain to obtain a target domain wind power prediction model. The wind power data distribution difference between the multi-source domain and the target domain is reduced through the Euclidean alignment method, the knowledge migration efficiency is improved, and the negative migration risk is reduced; setting a migration weight factor for each source domain based on inter-domain MMD, so that the model preferentially learns high-correlation source domain wind power characteristics; and the multi-source domain deep migration learning model is fused into a dynamic weight module, so that low-efficiency migration is avoided.

The invention discloses an equipment agent processing program conversion method and device based on cooperation of a large model and a small model. An equipment agent is constructed by analyzing control systems, kinematics structures and process constraint information of a source machine tool and a target machine tool, and line-level instructions are abstracted into system-independent unified semantic representation based on a multi-system numerical control corpus. And performing grammar analysis and semantic mapping on the source program, and generating a conversion task in combination with grammar rules, kinematics accessibility and process security constraints of a target machine tool. A small model is finely adjusted through supervised learning to realize basic generation capability, a large model is introduced as a patch evaluator, candidate results are jointly scored from different dimensions, knowledge migration is realized based on improved near-end strategy optimization, and a small model generation strategy is continuously optimized. And finally, high-quality results are screened through confidence degree sorting, cross-system G / M instruction and multi-axis track generation and verification are completed, and high-precision automatic conversion of numerical control programs is achieved.

The invention discloses a hyperspectral image classification method for cross-domain small sample learning based on diffusion enhancement prototype knowledge distillation, and belongs to the technical field of hyperspectral image processing. The method comprises the following steps: extracting a neighborhood data cube, aligning spectrums, dividing a support set and a query set, and applying a mask and enhancing noise; executing domain adversarial denoising and reconstruction tasks, aligning feature distribution, and outputting a pre-training encoder; decoupling features, capturing spectrum-space global and local dependency relationships, and calculating similarity between a query set and a category prototype; constructing a distillation framework to realize knowledge migration; optimizing model parameters, and introducing a signal-to-noise ratio to enhance loss suppression noise; and performing feature extraction by using the optimized student model to generate a hyperspectral image classification result. According to the method, the problems of domain offset, intra-class feature dispersion, inter-class boundary fuzziness, noise interference and the like are solved, and the classification accuracy in a small sample scene is improved.

The invention discloses a deep reinforcement learning optimization method for injection molding process parameters, and belongs to the technical field of intelligent manufacturing. The method comprises the following steps: constructing a dynamic causal graph network through information entropy flow analysis and transfer entropy calculation, and revealing a causal relationship and time delay characteristics among process parameters; manifold learning is adopted to map a high-dimensional parameter space to a low-dimensional manifold, and Riemannian metric guide optimization search is constructed based on the quality gradient; generating enhanced state representation fusing causal association and manifold geometric information; identifying a production element state and selecting a corresponding optimization strategy; a geodesic line is planned in a manifold space to obtain an optimal parameter adjustment path; historical experience is utilized through memory retrieval and case adaptation; cross-task knowledge migration is realized; adopting a depth deterministic strategy gradient algorithm to optimize the decision; and online learning is realized through elastic weight consolidation. According to the method, the problems of black box decision, slow convergence, difficulty in knowledge reuse and the like in the prior art are solved, the optimization efficiency and the interpretability are improved, and the method has the capability of quickly adapting to new tasks.

The invention relates to the field of artificial intelligence and medical image processing, in particular to a mammary gland pathology visual model establishing method based on multi-model fusion and combined distillation, which comprises the following steps: performing tissue segmentation and dyeing normalization on a full-slice image; inputting the image blocks into a pre-training teacher model of a plurality of freezing parameters in parallel to extract high-dimensional features, and generating unified enhanced features through a learnable feature fusion network; constructing a student model, and performing end-to-end training by using a joint loss function including feature simulation, logic output distillation and multi-task supervision; and connecting a plurality of task specific prediction heads to the student model, and realizing full-slice-level multi-task diagnosis and treatment prediction through an aggregation strategy. According to the technical scheme, efficient knowledge migration and multi-task cooperation can be achieved, and the accuracy, generalization ability and reasoning efficiency of mammary gland pathology analysis are remarkably improved.

The invention provides an online rail fault diagnosis method and system based on knowledge transfer learning, and belongs to the technical field of crossing of intelligent monitoring and artificial intelligence of railway infrastructures. The method comprises the steps that S1, a server trains a model framework through a source domain data set to obtain a teacher diagnosis model; based on the target domain data set, training the teacher diagnosis model by adopting a mixed training strategy to obtain a student diagnosis model; s2, acquiring real-time multi-modal monitoring data from a target domain line by the edge computing equipment, compensating to obtain corrected data, and dynamically selecting the most important feature subset from the corrected data to form a simplified feature set; and S3, inputting the simplified feature set into a student diagnosis model to obtain a student fault diagnosis result and confidence thereof, and introducing a D-S evidence theory to generate a student fault diagnosis report. The method has the advantages that the accuracy, the real-time performance, the cross-domain adaptability and the overall system safety of railway track fault diagnosis are greatly improved.

The invention belongs to the technical field of artificial intelligence and multi-modal emotion calculation, and discloses a multi-modal emotion recognition method based on a Mama state space model and cross-modal self-distillation. Through the organic combination of the efficient sequence modeling capability of the Mamba state space model and the knowledge sharing mechanism of cross-modal self-distillation, the advantages of the state space model in the aspects of time sequence modeling and calculation efficiency are fully played, and meanwhile, the limitation of a single model architecture is made up through a cross-modal attention mechanism; the technical bottlenecks of an existing multi-modal emotion recognition method in the aspects of long sequence processing efficiency, cross-modal information fusion and knowledge transfer sufficiency are effectively solved, and an efficient and reliable technical solution is provided for further development and practical application of the multi-modal emotion recognition technology.

The invention discloses a structure cross-domain damage identification method based on a pulse graph neural network. The method is suitable for structure health monitoring in the fields of civil engineering, mechanical equipment and the like. According to the method, structure dynamic response signals are collected through a sensor array, a graph structure is constructed, the graph structure is converted into space-time pulse graph data through pulse coding, and the space-time pulse graph data are input into a pre-trained pulse graph neural network for feature extraction and recognition. According to the method, the low power consumption and event-driven characteristics of the spiking neural network are creatively combined with the spatial topology modeling capability of the graph neural network, and a domain adversarial transfer learning mechanism is introduced, so that the problem of generalization of the model in cross-domain scenes of different structures, different environments and the like is effectively solved. According to the method, accurate positioning and quantitative evaluation of damage can be realized, and the method is particularly suitable for knowledge migration and long-term online monitoring from a laboratory model to a real structure.

The invention relates to the technical field of geotechnical material local modeling, in particular to a geotechnical material data driving local modeling method and device based on a knowledge migration strategy, and the method comprises the steps: building a first mapping data pair based on the stress-strain physical quantity of a test curve sampling point, and building a high-fidelity data set of a geotechnical material; constructing a second mapping data pair based on the stress-strain state of the simulation curve sampling point and the intermediate physical quantity label, and establishing a low-fidelity data set of the geotechnical material; pre-training the target data-driven constitutive model by using the low-fidelity data set to generate an initial geotechnical material data-driven constitutive model; and driving the model to be fine-tuned by using the high-fidelity data set for the initial geotechnical material data, and generating final geotechnical material data to drive the model. Therefore, the problem of insufficient generalization ability and prediction precision of a data-driven model due to limited training data and unpredictability of a stress path in the prior art is solved.

The embodiment of the invention provides a multi-level federal multi-mode large model privacy protection enhancement method and electronic equipment, and belongs to the technical field of intelligent traffic systems. According to the method, a'end-edge-cloud 'three-level federated learning architecture is constructed, firstly, local differential privacy disturbance is applied to local multi-mode traffic data at a traffic terminal node, and an end-side local model is trained; then aggregating a plurality of end side models on an edge server, and generating a personalized small model reflecting regional characteristics; then cooperatively training a plurality of personalized small models in a cloud center server to generate a global multi-modal large model; finally, knowledge of the global large model is fed back to a lower-level model through knowledge migration, and a continuously evolved iterative closed loop is formed. According to the method, privacy protection is carried out at a data source, so that original sensitive data is ensured not to go out of the local, the problems of data islands and privacy leakage in traffic large model collaborative training are effectively solved, and efficient and credible collaborative modeling is realized on the premise of ensuring data sovereignty.

The invention belongs to the technical field of pipeline safety assessment, and discloses a multi-dimensional information interconnection and autonomous evolution collaborative pipeline full-state safety assessment method. And capturing a high-order relationship of data through double hypergraph reasoning of the instance-level hypergraph and the modal-level hypergraph to realize efficient interconnection. According to the method, mode-level and instance-level hypergraph information features are extracted through hypergraph information propagation, high-order correlation is mined through double-graph information aggregation, cross-mode and cross-instance consistency information and exclusive information are output after feature recombination, multi-dimensional data deep fusion is promoted, and high-quality data support is provided for follow-up pipeline full-state safety assessment. A two-stage autonomous evolution mechanism of intra-class progressive calibration and inter-class knowledge migration is respectively adapted to slight fluctuation and significant change scenes of the deep sea environment: precise adaptation of environment perturbation is realized through dual-branch feature extraction and dynamic weight adjustment in a domain; model parameter dynamic optimization is completed between domains through spatial-temporal feature clustering and cross-domain knowledge migration, and dynamic environment self-adaption can be achieved without manual intervention.

The invention provides a cross-well-area drilling parameter migration method and system based on migration learning, and the method comprises the steps: firstly obtaining a drilling parameter historical data set, a scene data set and an operation effect data set of a source well area, and an initial drilling scene data set and an operation demand data set of a target well area; constructing a cross-well-area drilling parameter transfer learning model comprising a source domain feature extraction layer, a domain adaptation layer, a knowledge transfer layer and a target domain parameter generation layer, inputting source well area data into the source domain feature extraction layer to extract feature vectors, and inputting the feature vectors and target well area scene data into the domain adaptation layer to adjust a parameter mapping matrix; inputting the vector after domain adaptation and source well area operation effect data into a knowledge migration layer to generate a target domain preliminary parameter feature vector, inputting the target domain preliminary parameter feature vector into a target domain parameter generation layer in combination with target well area operation demand data to generate an adaptive parameter set, and finally transmitting the parameter set to a target well area drilling control system to execute drilling operation. And the accuracy and efficiency of cross-well-area drilling parameter migration are improved.

The invention relates to the technical field of model lightweight, and particularly discloses a deep neural network model optimization method based on hierarchical reinforcement learning and multi-agent collaborative distillation, and the method comprises the steps: building a structured pruning searcher based on an ABC algorithm, constructing a pruning combination reduction strategy dynamic artificial bee colony pruning algorithm, and carrying out the optimization of a deep neural network model. Performing fitness evaluation to guide a search process, and outputting an optimal pruning network structure under resource constraint; establishing a staged distillation architecture and a multi-dimensional hierarchical loss function, and realizing smooth and progressive knowledge transmission between the teacher model and the assistant model; a fine-grained quantization scheme based on parameter classification is designed, differential bit widths are configured for weights, batch normalization parameters and activation output respectively, a quantization perception training loss function fusing a hardware delay look-up table model is constructed, hardware perception joint fine tuning of the network weights and quantization parameters is achieved, and the quantization precision of the network weights and the quantization parameters is improved. Therefore, the effect of remarkably improving the model compression efficiency on the premise of keeping the precision is achieved.

A text sequence recommendation method and system based on a large language model is disclosed, belonging to the technical field of recommendation algorithms. The method includes: a data preprocessing stage, a large language model pre-training stage, a sequence model fine-tuning stage and a matching stage. According to this disclosure, a large language model is introduced into a text sequence recommendation task, so that text can be better modeled by utilizing rich pre-training corpus of the large language model; meanwhile, sequence modeling is performed on the text, the capability of sequence recommendations modeling in a large model is activated, an ID-based recommendation paradigm in a traditional recommendation algorithm is eliminated, and recommendation task learning processing is better performed in a cold start scenario and a knowledge transfer scenario; and finally, a recommendation result is finally optimized by a sequence model.

The invention provides a Riemannian graph word segmentation device for structure knowledge migration, which belongs to the field of graph basic models, and comprises a geometric vocabulary sampling module used for sampling input graph data to obtain geometric vocabularies; the geometric vocabulary encoding module is used for respectively mapping geometric vocabularies into corresponding constant curvature Riemannian spaces for coordinate encoding according to different structural modes of the geometric vocabularies to obtain space coordinates corresponding to the structural modes of the geometric vocabularies, discretization is carried out by using the Riemannian quantization module to obtain quantization marks, and finally, the geometric alignment decoding module is used for decoding the geometric vocabularies. And fusing the quantitative marks in different Riemannian spaces to obtain a unified graph structure representation capable of supporting structure knowledge migration. The problems that in the prior art, due to the fact that a curvature selection mechanism of a local form of a graph structure is lacked, a word segmentation device cannot dynamically adapt to geometric characteristics of the structure, and structural representation aliasing exists, the graph structure coding accuracy is insufficient, and the cross-domain knowledge migration generalization ability is low are solved.

The invention discloses a mechanical equipment residual life prediction method based on multiple expert models, and belongs to the technical field of mechanical equipment. Comprising the steps of receiving an original mechanical vibration signal, performing preprocessing, constructing a residual life prediction model of the mechanical equipment, training the residual life prediction model of the mechanical equipment based on a preset data set, inputting a mechanical vibration signal to be detected into the trained residual life prediction model of the mechanical equipment, and performing fault category prediction of the mechanical equipment. According to the method, the accuracy, generalization and practicability of residual life prediction of mechanical equipment are remarkably improved, and through multi-scale time-frequency domain feature fusion and a double attention mechanism, the identification capability and noise immunity of the model to a complex fault mode are enhanced; the domain adaptive routing network realizes cross-device efficient knowledge migration, greatly improves the prediction performance on unseen devices, has high precision and high real-time performance, and provides reliable technical support for industrial predictive maintenance.

The invention discloses a model generalization ability optimization method and system based on hybrid experts, and aims to solve the problems of weak generalization ability, difficulty in knowledge migration, low semantic alignment efficiency and the like when a multi-modal large model performs fine adjustment on downstream tasks in a new field. The method comprises the following steps: embedding a hybrid expert module in a low-rank adaptive fine tuning framework of a self-attention module in the visual encoder, introducing a heterogeneous hybrid convolution structure, innovatively realizing transverse and longitudinal independent scaling through bilinear interpolation in the convolution structure, supporting a multi-scale convolution kernel, and realizing dynamic combination through an MoE framework; constructing a cross-modal expert adapter of channel perception, and adopting dynamic channel recombination, cross-modal conditional convolution and weighted spatial feature aggregation methods; according to the method, the problem of local semantic confusion can be well solved, meanwhile, semantic fusion of image features and text features is enhanced, more effective spatial information and priori knowledge are injected into the model, and therefore the generalization of the model is improved.

The invention provides an anticancer drug collaborative prediction method MultiFusion Syn based on multi-scale feature fusion, and relates to the technical field of anticancer drug collaborative prediction. According to the method, synergistic effect prediction is converted into a multi-scale feature fusion classification task: a pre-trained graph neural network DGCL is utilized to extract drug fine-grained structure features, ChemBERTa is combined to obtain semantic features, and adaptive fusion is carried out through an attention mechanism; meanwhile, a residual network is adopted to extract biological characteristics of the cell line, drug-cell bidirectional interaction is constructed by means of double-end cross attention, and fine fusion is performed through a gating network; the model introduces a pre-training encoder and realizes knowledge migration in combination with parameter freezing. On a reference data set, MultiFusionSyn is evaluated by indexes such as AUC, PRAUC, ACC and BACC, and the performance of MultiFusionSyn is superior to that of an existing advanced method through verification of an independent test set.

The invention relates to the technical field of data detection, and discloses a gold ore vein measuring and positioning method based on multi-data coupling, and the method comprises the steps: building an omnibearing monitoring basis through multi-source data collection, building a separation trend and abnormality through a dynamic baseline, and determining a real causal path through the recognition of a causal structure. Training data are enriched through anti-fact sample expansion, graph network modeling is adopted to fuse spatial topology and a causal mechanism, incremental online updating is implemented to maintain model adaptability, physical mechanism verification is performed to ensure mechanical rationality, finally, an adjustment scheme is output through risk prediction decision, and a knowledge migration multiplexing mechanism is established to realize cross-project application. According to the complete technical chain, cognitive upgrading from surface data association to a deep causal mechanism is achieved, settlement prediction is converted into scientific decision from experience judgment, and prediction accuracy, interpretability and generalization ability are improved.

The invention discloses a real-time medical image segmentation method and system, and the method comprises the steps: carrying out the pre-training processing of a basic model in a self-supervised learning mode, and enabling the basic model to learn domain invariant features from inter-frame dependence based on the time series data input of a medical image sequence; the multi-scale feature knowledge of the basic model is migrated to the lightweight detection model through feature distillation operation, and the feature distillation operation is configured to be based on a feature alignment mechanism, and fine features are not lost during knowledge transfer; and automatically generating segmentation prompt information by using a lightweight detection model, inputting the prompt information as spatial guidance into a segmentation model, driving the segmentation model to output a pixel-level result, and constructing an end-to-end flow workflow without manual intervention. According to the invention, through organic combination of self-supervised pre-training, feature distillation and automatic prompt generation, end-to-end automatic segmentation is realized.

The invention provides a teacher-query guide type compression optimization training method and system based on open domain questions and answers and a query method and system. The compression optimization training method comprises the steps that candidate paragraphs are obtained according to user queries, each query is matched with the corresponding candidate paragraphs, the candidate paragraphs are segmented into sentence sets, and training units are formed; the teacher model performs data distillation on the training unit based on a preset requirement to generate a distillation sample set; the method comprises the following steps: respectively coding queries and sentences in a distillation sample set into vector representations through a predefined coding function, measuring the similarity of the queries and sentences, and training a student model by taking the similarity as a supervision signal to obtain a trained student model; through a teacher model knowledge migration and query focusing mechanism, the accuracy and efficiency of a student model in an extraction type compression task are improved, redundant information interference is reduced, more compact and highly related context input is provided for a retrieval enhancement generation framework, and the method is suitable for an efficiency-sensitive open domain question and answer scene.

The invention relates to a multi-scale feature recognition method for power equipment in the technical field of intelligent detection of the power equipment. Aiming at the problems of high calculation complexity, insufficient real-time performance and weak multi-scale target identification capability of a traditional electric power detection model, the method is optimized through the following technical scheme: replacing a self-attention mechanism in a Grouping-DINO model with a Comba recursive structure, constructing a lightweight student model, and constructing a multi-scale target identification model; knowledge migration from a teacher model to a student model is realized through a multi-task distillation loss function, a four-scale feature pyramid network is added behind a backbone network of the student model, and feature extraction and key area focusing of a multi-scale target of power equipment are enhanced in combination with a channel self-attention weighting mechanism. The method effectively maintains the cross-modal feature fusion capability and detection precision of the model while remarkably improving the reasoning speed and reducing the model volume, and is particularly suitable for real-time detection of power equipment in edge calculation scenes such as unmanned aerial vehicle inspection.

The invention belongs to the technical field of unmanned driving environment perception, and provides a night semantic segmentation method and system based on passive multi-level collaborative distillation, and the technical scheme is as follows: firstly, initializing a teacher network and a student network based on pre-training model parameters under a normal illumination condition; then, teacher network parameters are fixed, and a teacher network is updated according to student network parameters in a momentum smooth propagation mode; then, based on a prediction result of a pre-training source model on the night image, selecting first K pixel points with the highest confidence coefficient to generate a pseudo tag; and finally, inputting the normal illumination image and the night image into a teacher network and a student network respectively, and realizing multi-level knowledge migration through structure perception level alignment, semantic consistency constraint optimization and frequency domain collaborative fusion. No night annotation data is needed, source domain original data does not need to be accessed, and the data cost and the privacy risk are remarkably reduced.

The invention relates to a high-efficiency federal feature distillation method based on an intelligent agent, and the method comprises the steps: firstly constructing a system which comprises clients and a server, and enabling the clients to enrich the content of knowledge migration through sharing features; then, an intelligent agent driven by reinforcement learning is introduced into a server of the system, the training environment is dynamically perceived according to the intelligent agent at the server side, the most effective feature subset and the weight thereof are adaptively selected for each training round, training of the client side and the server is further continuously optimized, and finally a trained local model is obtained; according to the method, an efficient, intelligent and practical collaborative learning solution is provided for the wireless edge network, dynamic adjustment between communication and computing resources and model performance is achieved, communication and computing overhead of the system can be remarkably reduced, and the best balance between knowledge migration depth and resource utilization efficiency is achieved.

The application relates to a large-scale road network traffic control method based on a deep reinforcement learning large model and belongs to the technical field of intelligent traffic control. The method comprises the following steps: perceiving real-time multi-modal road network information including urban road intersections, highway entrance ramps and emergency lanes, generating a space-time fusion representation vector representing the current traffic network state by fusing a space graph construction method and a time series embedding method; splicing the space-time fusion representation vector and a historical state memory to serve as input, performing state feature distillation on a backbone network of a pre-trained large language model to enhance state representation, and outputting a traffic control decision through a policy network with a hierarchical action space; and guiding and optimizing the training process of the deep reinforcement learning algorithm through cross-modal knowledge transfer and a progressive curriculum learning strategy to improve the model training efficiency and generalization ability. The application improves the generalization performance and accuracy of the control strategy while ensuring real-time response speed.

The present application relates to the field of data detection, in particular to a network traffic anomaly detection method based on polymeric quasi-state distillation. The present application performs protocol level analysis on traffic through a multi-branch feature extraction network, and generates a unified representation vector through a cross-layer fusion mechanism; a first anomaly score is calculated based on a protocol-aware weighted adversarial soft contrast mechanism; a heterogeneous teacher model is constructed and dynamically weighted and aggregated, and a student model learns to generate a second anomaly score through knowledge distillation; the student model, the teacher model and the rule detector are combined into a heterogeneous redundant detection pool, a detector is dynamically selected and adaptive disturbance is applied to obtain a third anomaly score; and the three types of anomaly scores are dynamically fused to output a detection result. The present application effectively solves the problems of protocol semantic fragmentation, weak boundary sample discrimination, single knowledge transfer and predictable defense path, significantly improves the detection accuracy, robustness and dynamic defense capability.

Systems and methods for non-contact arterial blood pressure waveform monitoring based on machine learning are provided. The method includes transmitting a millimeter wave signal waveform with a millimeter wave radar and receiving a reflected signal from the chest of a subject; extracting complex value millimeter wave IQ data from the received reflection signal; guiding the complex valued millimeter wave IQ data with a beamforming-based data enhancement module to form a target signal beam; and estimating an arterial blood pressure waveform from the target signal beam using a millimeter wave-arterial blood pressure waveform converter. During a training phase, the beamforming-based data enhancement module is configured to generate millimeter wave signal beam data from different angles for training the millimeter wave-arterial blood pressure waveform transducer; and a cross-modal knowledge migration module is utilized to generate a teacher model so as to jointly supervise training of the millimeter wave-arterial blood pressure waveform converter.

This application relates to a method, apparatus, device, and medium for few-sample SAR target recognition. The method includes: dividing SAR samples into support and query sets based on a meta-learning framework; converting the original SAR image into a semantic map and an attribute scattering center topology map; extracting three types of features and calculating class prototypes through three parallel feature branches; obtaining the confidence score of each branch using the Softmax function; assigning teacher and student branches according to the confidence score; minimizing KL divergence to achieve dynamic knowledge transfer; weighted fusion of predicted log odds followed by Softmax output; and backpropagation to optimize the network. This method, through multi-prototype fusion and interactive distillation, mitigates the problems of speckle noise and high intra-class variance in SAR images, maintaining high robustness and generalization ability even in extreme data-scarce scenarios.