850 results about "Interpretability" patented technology

The application belongs to the technical field of artificial intelligence, and specifically discloses a power customer service model optimization method and system based on artificial intelligence, which obtains a support set and a query set for large model small sample learning, uses the support set and the query set to perform meta-training on a large model, retrieves an auxiliary set to perform inference test on the pre-trained large model for power business scene problem instances, and then performs reinforcement learning on the pre-trained large model based on a comprehensive strategy reward value according to the inference steps and the inference result obtained through the test, so as to obtain a strategy optimized large model to infer and optimize the solution to the power business scene problem in actual application. The application uses small sample learning technology to realize fine-tuning of the inference ability of the large model, and uses comprehensive strategy feedback to perform reinforcement learning, thereby improving the explainability of the inference process of the large model and the accuracy of the inference result, and providing more efficient decision support for power customer service question answering of the large model.

This application provides a joint emotion recognition method and system based on action unit-driven attention, relating to the fields of computer vision and affective computing. The method includes: inputting a feature map into an action unit (AU)-driven attention branch, processing it through a sigmoid activation function to obtain an activation vector, and inputting this vector into a learnable mapping layer; projecting the AU information back into the spatial geometric space to generate a single-channel spatial attention map, which is then multiplied element-wise with the feature map to obtain a weighted feature map; determining continuous values ​​for discrete emotion category probabilities, valence, and arousal; constructing a total loss function; and training the model using the acquired AffectNet and Aff-wild2 datasets to determine the target model for emotion recognition. This application achieves end-to-end joint optimization of AU detection and emotion recognition, significantly enhancing the interpretability and generalization ability of the model while improving emotion recognition accuracy, enabling it to better adapt to the emotion recognition needs in complex scenarios.

The application relates to an excitation system fault recording and event record analysis and diagnosis method and system, belonging to the field of excitation systems. The method comprises collecting recording files and event sequence records of the excitation system; performing multi-domain feature extraction based on the recording files to obtain a multi-domain feature tensor; performing space-time causal structure learning based on the multi-domain feature tensor and the event sequence records to output a causal adjacency matrix, a causal diagram and a time delay matrix; performing double-channel interpretable fault classification based on the multi-domain feature tensor, the causal diagram and an event time tag list E in the event sequence records to output a fault type label M and an attention space-time heat map; performing counterfactual causal tracing to obtain a root cause variable set and a causal propagation path, and outputting a diagnosis report. The application realizes intelligent diagnosis of the excitation system with signal analysis capability, causal reasoning capability and diagnosis interpretability.

The present application discloses a bearing fault diagnosis method and system based on knowledge enhancement and multi-path distillation, which relates to the fields of intelligent operation and maintenance and industrial equipment health management. The method includes obtaining sensor signals and visual image data of the bearing operation; based on an asynchronous dual-channel architecture, correspondingly extracting signal features of the sensor signals and image features of the visual image data, and performing time synchronization on the signal features and the image features; using a multi-modal bottleneck Transformer module to fuse the synchronized signal features and the synchronized image features; based on a maintenance knowledge graph dynamically constructed from a bearing maintenance manual, combining a text generation model to map the fused features to a semantic space and generate a fault diagnosis report. The present application can improve the recognition accuracy, real-time performance and interpretability of diagnosis results of bearing faults.

The application discloses an air dynamic test platform vacuum equipment leakage fault diagnosis method and system, aiming at solving the problems that the existing method is difficult to adapt to the complex time-varying characteristics of mixed gas, data is sparse, and lack of interpretability. The method first collects time series pressure data through a conventional pressure sensor, uses the ideal gas state equation as a physical mechanism model, deduces a leakage rate calculation equation, and extracts key characteristics of instantaneous leakage rate and cumulative leakage rate; then a BP neural network is constructed as a data-driven model, taking the leakage rate as input, and outputting normal and abnormal judgments after training with historical data. The application innovatively integrates physical mechanism and machine learning, without the need for special detection equipment, ensuring the interpretability of the diagnosis results, improving the diagnosis accuracy and reliability under small sample and variable working conditions, and realizing high-precision, low-cost air tightness online monitoring and early warning.

The application discloses an ancient Chinese intelligent translation method and system based on field adaptive retrieval enhancement, and belongs to the field of ancient Chinese translation. The method comprises the following steps: S1, constructing an ancient Chinese translation basic resource library; S2, determining a target field of input ancient Chinese original text and field determination confidence; S3, performing double-source fusion retrieval based on the target field and the field determination confidence; S4, determining high-confidence ancient and modern interpretation information; S5, extracting the origin books, specific chapters, original texts and modern Chinese translations corresponding to the allusions, as complete allusion tracing information; S6, constructing a field adaptive structured translation prompt; S7, generating a modern Chinese translation corresponding to the input ancient Chinese original text; and S8, integrating an interpretable translation evidence chain. The ancient Chinese intelligent translation method and system based on field adaptive retrieval enhancement can realize the full-link interpretability, verifiability and traceability of the translation process.

This invention relates to the interdisciplinary field of assisted reproductive technology and artificial intelligence, specifically to a system and method for constructing an embryonic developmental trajectory map that integrates morphological features. By deeply integrating static morphological features with dynamic morphodynamic parameters, a more comprehensive panoramic map of embryonic development with richer information dimensions is constructed, overcoming the one-sidedness of single-dimensional assessment. Automated analysis of the map using an AI model reduces the influence of subjective human factors, enabling the discovery of subtle developmental patterns and potential differences that are difficult to discern with the human eye, significantly improving the accuracy of selecting high-quality embryos. The provided developmental trajectory map allows embryologists to intuitively examine the entire developmental process of the embryo and the performance of each key node, enhancing the interpretability and credibility of the assessment process. Through more scientific embryo selection, it is expected to improve embryo implantation and clinical pregnancy rates, and reduce multiple pregnancy and miscarriage rates.

This invention discloses a drought early warning method based on energy flux, causal relationship analysis, and machine learning. The method includes the following steps: S1, acquiring energy flux and drought indicators, determining the optimal lag time through causal reasoning, and constructing a multi-order lag feature set; S2, establishing a tree model, using regression / kernel methods and a time series model candidate set, optimizing hyperparameters through particle swarm optimization, integrating two layers in a stacked manner, adaptively optimizing the performance of comprehensive regression and event recognition, and outputting a predicted sequence; S3, setting multi-level early warning rules according to drought thresholds, mapping drought levels, and evaluating effectiveness through statistical precision and recall; S4, calculating contribution using an additive feature attribution algorithm, identifying nonlinear thresholds to form sensitivity analysis, and improving interpretability. This invention achieves a 7-11 month early warning prediction of drought based on energy flux, with a drought early warning recall rate of 66.67%-75.86%, significantly improving the accuracy and interpretability of drought early warning.

The application belongs to the technical field of artificial intelligence medical treatment, and discloses a medical health service system and method based on a multi-granularity semantic analysis and knowledge reasoning engine, which comprises the following steps: through a multi-granularity semantic analysis module, coarse-grained intention recognition and fine-grained medical entity state extraction are performed on the unstructured input of a user; a dynamic probability reasoning engine starts two reasoning paths in parallel: a generative reasoning path generates a group of candidate diagnosis hypotheses by using an LLM, and a symbolic reasoning path performs multi-hop probability reasoning on a unique probability knowledge graph to calculate another group of candidate diagnosis paths and their cumulative probabilities; a reasoning fusion and verification module cross-verify the results of the two paths, and only when the hypothesis of the LLM is verified by the high-probability path of the PKG, a final health service response containing a conclusion and an interpretable path is generated. The application significantly improves the accuracy, reliability and transparency of automated medical consultation services.

A deep learning-based method and system for statistical analysis of fluorescence proportion information in microfluidic droplet arrays, belonging to the field of ddPCR detection technology, includes the following steps: S1 acquiring a three-dimensional fluorescence volume image of the droplets and performing preprocessing; S2 constructing a DropletSegNet model; S3 pre-training the DropletSegNet model constructed in S2; S4 using the three-dimensional fluorescence volume image to be tested as input to the DropletSegNet model, and transforming the original output of the DropletSegNet model into droplet analysis results with physical meaning and biological interpretability. A droplet stacking topology decoupling module is designed, modeling the adhesion decoupling as an edge classification problem on a graph structure, and using a graph neural network (GCN) for structured post-processing. This achieves accurate decoupling and segmentation even in cases of tight droplet stacking or adhesion, improving the accuracy of nucleic acid concentration calculation and positive proportion statistics.

Embodiments of the present application disclose an intention recognition method and device, electronic equipment and medium, relating to the technical field of artificial intelligence. A specific embodiment of the method comprises: constructing an intention knowledge graph reflecting the association relationship between historical intention labels based on at least one historical intention label in historical dialogue text; performing semantic topic division on at least one historical intention label in the intention knowledge graph to obtain at least one historical intention label set; performing semantic similarity matching on target dialogue text based on the at least one historical intention label set to obtain at least one candidate intention label associated with the target dialogue text; and performing intention recognition on the target dialogue text based on the at least one candidate intention label to determine the target intention of the target dialogue text. The intention label system is dynamically expanded, the intention semantic association is utilized to improve the interpretability of the recognition result, and the complex customer intention recognition requirement is adapted.

The application discloses an osteoporosis risk grading prediction method based on clinical prior logic gating, acquires multi-modal clinical data of a patient to obtain a standardized feature vector group, constructs and trains an osteoporosis risk grading prediction neural network model, including a double-flow feature coding module, a logic gating fusion module and a cascaded classification prediction module, inputs the standardized feature vector group after splicing processing to the trained neural network model, and outputs an osteoporosis risk grading prediction result. The application converts the diagnosis logic in the clinical guideline into an attention gating mechanism in the neural network by designing a double-flow architecture, and combines a screening-grading cascaded strategy, so that the existing structured data (demography, test single, medical history) in the hospital information system can be directly used to realize low-cost, high-precision and clinically interpretable osteoporosis risk grading without additional image examination.

This invention relates to the field of industrial equipment fault diagnosis technology, specifically to a knowledge graph-embedded method and system for causal tracing of industrial equipment events. The method includes the following steps: collecting multi-source heterogeneous event data during the operation of industrial equipment; preprocessing and extracting features from the collected multi-source heterogeneous event data to obtain event feature vectors. This invention constructs a knowledge graph in the industrial equipment domain, structurally representing the complex mapping relationships between entities such as equipment type, fault type, fault cause, and fault result, providing rich semantic prior knowledge for subsequent causal inference. Compared to traditional data-driven fault diagnosis methods, the knowledge graph introduced in this invention can encode domain experts' in-depth understanding of equipment fault mechanisms, making the tracing results more accurate and interpretable.

This application discloses an adaptive control method and electronic device for PVD and micro-arc oxidation equipment. The method includes: constructing a parameter inversion model characterized by a forward mapping relationship between prior distributions of process parameters and physical constraints; establishing a causal topology based on the mapping relationship, and identifying anomaly types by calculating the joint probability density of the two hypotheses; executing different strategies according to the anomaly type: shielding faulty sensors and reconstructing signals when measurement anomalies occur, and calculating parameter corrections through online optimization when mechanistic anomalies occur; performing safety verification and smoothing filtering on the corrections based on safety constraint envelopes; and generating an interpretable report containing root causes, contribution levels, and maintenance guidelines. This application achieves accurate anomaly type differentiation and robust adaptive control through physical-data decoupling modeling and causal reasoning, significantly improving the stability, safety, and interpretability of processes in complex industrial environments.

The application discloses a fan control parameter online optimization method and system based on digital twinning, which inputs real-time SCADA data into a fast digital twinning model, performs millisecond-level state estimation to meet the real-time response requirement of the control system. At the same time, the slow digital twinning model is driven by buffer data for deep calibration, a high-fidelity state log containing physical mechanisms is generated, and the knowledge distillation technology is used to migrate the physical knowledge of the slow model to the fast model for periodic correction of the fast model. Finally, the modified fast model is used for rolling optimization under the model predictive control framework to output the optimal control sequence. The scheme effectively overcomes the defects of the traditional physical model that the calculation time is too long, and the pure data-driven model that is easy to drift and lacks physical interpretability, while ensuring the millisecond-level response speed, and improving the physical consistency of model prediction and the robustness of long-term operation.

The application discloses a specific field pipeline task type dialogue system based on an LLM, which comprises an inquiry subsystem and an answering subsystem, wherein: the inquiry subsystem parses and guides a user to supplement key information of a question according to user input task text, the answering subsystem summarizes key information of a dialogue task, and performs vector matching through a built-in vertical field local vector database to generate a final answer to the dialogue task. The application uses a large language model to realize natural language understanding, dialogue state tracking and natural language generation modules in a task type dialogue system pipeline, simultaneously realizes a rule matching-based strategy learning module, understands and replies to user input text, and combines the content of a local knowledge base, so that the whole process is more interpretable, the illusion problem of the large language model is reduced, only a small amount of sample prompt learning is needed, a large amount of data training or fine tuning is not needed, and the workload during field migration is reduced.

The present application relates to the technical field of artificial intelligence and recommendation system, in particular to a kind of dynamic interaction perception and multilayer semantic compression generative recommendation method, comprising the following steps: obtaining the original attribute information of article and carrying out semantic extension, generating enhanced article description text;The historical interaction sequence of user is obtained, and the historical interaction sequence is carried out time feature extraction, multi-factor weighting and gate memory state update, to generate the dynamic preference vector of user;The historical interaction sequence is divided into multiple blocks, and the block-level abstract of each block is generated using a large language model and recursively compressed to obtain a global preference abstract;Question and answer samples are constructed, and a large language model is trained based on the question and answer samples to obtain the trained large language model.The present application improves the modeling capability of the recommendation system for user interest evolution through dynamic interaction perception and multilayer semantic compression, reduces long sequence input redundancy, improves the inference efficiency of the large model, and enhances the accuracy and interpretability of the recommendation results.

The application discloses a rail transit passenger flow prediction method and system based on a traffic interpretable large model, and the method comprises the following steps: constructing a traffic field knowledge graph, constructing an input instruction containing a cause-effect logic guide based on obtained historical passenger flow data and external environment information; triggering the thinking chain reasoning of a preset large language model based on the input instruction to generate reasoning text containing a cause-effect logic chain, and performing fact consistency verification and correction on the reasoning text through the knowledge graph to obtain verified cause-effect logic information; fusing the cause-effect logic information and the spatiotemporal characteristics of the historical passenger flow data to obtain fusion characteristics; and outputting a passenger flow prediction value of a future period and a corresponding natural language explanation report based on the fusion characteristics. Through the knowledge graph, the thinking chain reasoning of the large language model is constrained, and the cause-effect logic information and the spatiotemporal numerical characteristics are fused, so that the collaborative optimization of the passenger flow prediction accuracy and the interpretability is realized.

The application relates to the field of natural language processing and decision intelligence, and particularly relates to a large language model reasoning method based on time series difference learning and rule enhancement, which is widely applied to automatic planning, intelligent question answering, embodied intelligence and the like. The method comprises task trajectory sampling, domain knowledge induction, domain rule extraction and large language model reasoning based on rule enhancement. For test task data, the most relevant historical task is matched based on vector retrieval, and a domain rule set corresponding to the task is obtained. The rule is rewritten in natural language by the large language model itself, so that the rule is more interpretable and adaptable. Finally, the optimized rule set is integrated into the large language model reasoning prompt text, so as to optimize the reasoning quality and stability.

This invention discloses a career path planning method and system based on LLM (Multi-Agent Learning). The method first generates a set of target constraints through intent and constraint parsing and requirement clarification. Then, it constructs individual ability profiles and job-specific quantitative profiles. Next, based on ability gaps, it generates task packages level by level along stage boundaries and supplements prerequisite dependencies, forming a set of candidate task packages. Finally, addressing the problem that traditional sorting or traversal methods struggle to balance search efficiency and path quality due to the large candidate task package combination space, complex dependencies, and coexisting time and risk constraints, an improved Monte Carlo tree search algorithm is employed to filter invalid search branches in real time, efficiently searching and outputting career goal paths with better comprehensive scores. This invention improves the search efficiency, path quality, and interpretability of career goal path generation under complex constraints and can be widely applied to career development scenarios such as internships, campus recruitment, social recruitment, and entrepreneurship.

The application discloses a Bloom cognitive hierarchy constraint-based achievement-oriented education diagnosis method and system. The method takes course outline text and student evaluation data as input, and realizes the automatic diagnosis of course goal achievement through three structural modifications in the large language model Transformer architecture: the logarithmic value of the cognitive hierarchy transfer matrix is embedded as a learnable bias item in the attention score calculation path to realize feature coding of cognitive hierarchy perception; the continuous differentiable relaxation technique is used to establish an end-to-end differentiable inference path for the course goal achievement weight; and the Bloom partial order constraint is coded as a training loss, so that the diagnosis result meets the cognitive hierarchy progressive relationship. The three modifications are optimized by a joint loss function, which significantly improves the diagnosis accuracy and promotes the interpretability and scientificity of the large language model diagnosis in the achievement-oriented education scene.

The application provides a large language model data mining interaction method and system based on traceable thinking chains, relates to the cross field of language model technology and ship industry capability data mining, and comprises the following steps: obtaining a data mining demand of a user and generating a thinking chain; monitoring a reasoning node of the thinking chain in real time and triggering a plug-in mechanism and user interaction when necessary; when the result is not as expected, locating a problem node through full-link tracing; and rolling back the thinking chain to the problem node and regenerating an execution link. The application improves the accuracy, interpretability and human-computer interaction efficiency of ship industry capability data mining, and enhances the adaptability of the system to complex business rules and exclusive knowledge in the ship field.

The present application relates to a bridge technical condition prediction method based on the fusion of graph neural network and time series modeling, belongs to the field of traffic infrastructure maintenance and artificial intelligence application technology, and solves the problems of insufficient utilization of space-time characteristics, lack of engineering logic constraints and rough prediction granularity of existing methods. The present application first collects historical technical condition data and pre-processes to obtain structured space-time sequence data; a double-branch feature extraction network extracts spatial correlation feature vectors and time series evolution feature vectors in parallel; a degradation trend prior feature vector is fused through a gating mechanism to generate a comprehensive feature vector; a double-branch continuous ordinal prediction head is used to output the initial continuous prediction score of the target bridge in the prediction year; finally, a time series consistency post-processing algorithm is used for logical constraint correction to generate the bridge technical condition grade prediction result. The present application effectively utilizes space-time characteristics and has strong engineering logic interpretability, and can realize continuous and accurate bridge technical condition prediction.

The application discloses a multi-modal feature collaborative generation analysis method and system for tumor survival prediction, and relates to the technical field of computer vision. The method comprises the following steps: acquiring multi-modal data and preprocessing to obtain a training data set; learning a first feature mapping relationship between multiple modes based on complete mode samples, and training a generator based on missing mode samples and the first feature mapping relationship to obtain a target generator, which generates virtual coding features of the missing mode; acquiring partial mode medical data of a target object, inputting the partial mode medical data of the target object into the target generator to generate virtual missing mode coding features of the target object; extracting at least one other mode coding feature from the partial mode medical data, fusing the virtual missing mode coding features and the at least one other mode coding feature, and determining a survival prediction result of the target object according to the fused features. The application improves the accuracy and interpretability of tumor survival prediction.

The present application relates to a kind of methods for realizing network content risk analysis based on key points driving, comprising the following steps: depth semantic analysis is carried out to original text, and risk key points are identified and extracted;According to the preset confidence threshold filtering low reliability key points, calling big model intelligent screening risk key points;Multi-source evidence online retrieval is carried out, and multi-source search results are integrated and risk tendency determination;Multi-dimensional risk research and judgment is carried out.The method for realizing network content risk analysis based on key points driving, device, processor and computer readable storage medium thereof of the present application are adopted, by reconstructing analysis process with risk key points as core, and comprehensively deciding by depth fusion local semantic knowledge and external real-time information, surpassing existing technology in precision, explainability and flexibility etc.Dimensions, provide solid technical foundation for constructing next-generation intelligent, efficient, reliable network content risk prevention and control system, and have important practical value and innovation.

The application relates to a kind of intelligent question and answer methods of digestive system disease based on knowledge graph, specifically to disease learning and question and answer field, through fusing deep semantic understanding and knowledge graph reasoning, the accurate response of complex multi-round question and answer in digestive system disease field is realized;It first aligns the user dialogue with the medical knowledge graph in depth, dynamically focuses on the relevant knowledge subgraph, then explores the logically coherent reasoning path in the subgraph through reinforcement learning, and finally generates traceable answers based on the path and dialogue history;This process ensures that the reply is not only accurate in information, clear in logic, but also has good interpretability;More importantly, by feeding back the quality of the generated answer to the optimization process of the reasoning path, it realizes the end-to-end collaborative optimization from semantic understanding, knowledge retrieval, logical reasoning to language generation, significantly improving the overall performance and practicality of the system.

This invention provides a task performance evaluation method based on a combination of hierarchical KAN and RBF networks. The method involves: constructing multiple KAN networks corresponding one-to-one with categories; each KAN network performs hierarchical nonlinear transformation and fusion of multidimensional feature variables of the same category, outputting low-dimensional feature variables; constructing an RBF network and inputting the low-dimensional feature variables output by each KAN network into the RBF network; the RBF network uses all low-dimensional feature variables as input, performs nonlinear mapping, and outputs the task performance evaluation result; finally, the transformation from a task performance evaluation index system to a task performance evaluation result is completed. This invention effectively fuses features from multiple categories using KAN networks to achieve feature dimensionality reduction, and uses the RBF method to quickly fit the performance model between input and output, improving the accuracy and efficiency of the evaluation. It addresses the shortcomings of existing technologies in modeling multi-category, multi-feature variable data, improving evaluation accuracy, computational efficiency, and interpretability.

The present application belongs to the technical field of hydrological prediction and water resources management, and specifically relates to a prediction method and system based on multi-scale decomposition of runoff sequence and dynamic weight reconstruction. The method first constructs a physical hydrological model based on meteorological driving data and generates a runoff simulation sequence; the simulation sequence is subjected to multi-scale decomposition by using variational mode decomposition, the decomposition parameters are adaptively determined by particle swarm optimization, and a plurality of mode components are obtained; a long short-term memory network is constructed to establish a mapping relationship between the contribution weights of the mode components, and dynamic weights varying with time and normalized are output; the mode components are weighted and reconstructed according to the weights, so as to realize deviation correction of the simulated runoff of the physical hydrological model on different time scale structures. In the prediction stage, the same decomposition is performed on the future runoff simulation sequence, and the trained model is directly used to output the runoff prediction result. The present application converts the runoff prediction problem into a dynamic weight distribution problem of multi-scale structure components, improves the prediction precision and migration ability while maintaining physical interpretability, and is suitable for scenarios such as basin runoff prediction, flood simulation, water resources scheduling and the like.

This invention discloses a method and system for detecting energy storage charging anomalies based on user switching behavior, belonging to the field of energy storage anomaly identification technology. Addressing the problems of existing technologies that only focus on the physical state of equipment, cannot cover soft anomalies, rely on fault labels, have weak generalization ability, and lack interpretability, this application identifies effective charging segments from electricity consumption time-series data through a dual-threshold hysteresis comparison mechanism. Then, it extracts basic state features, strategy template compliance features, and behavior profile compliance features from the effective charging segments to construct a high-dimensional vector. This high-dimensional vector is input into a reconstruction autoencoder model trained on normal data. The anomaly degree is calculated through the reconstruction residual. Based on the contribution of each feature factor in the input vector to the anomaly degree, a fault index is performed using a rule base to determine the root cause of the fault. This achieves accurate detection and interpretable diagnosis of multiple types of soft anomalies in energy storage charging, eliminating the need for fault labels and significantly improving anomaly detection efficiency and accuracy.

The application provides an interpretable aerodynamic coefficient prediction method and system based on dynamic weighting. The method comprises: obtaining a flight state parameter feature dataset composed of a plurality of sample points and a corresponding aerodynamic coefficient label dataset, performing normalization processing on the feature dataset to obtain standardized sample features; training a group of heterogeneous machine learning models in parallel based on the standardized sample features and the label dataset, and obtaining unbiased prediction values of each model for each sample point through cross-validation; for each sample point, calculating and fusing three types of sample-level evaluation weights based on the unbiased prediction values of each model corresponding to the sample point to generate dynamic fusion weights of each model for the sample point; for each sample point, performing weighted summation on the unbiased prediction values of each model for the sample point by using the dynamic fusion weights corresponding to the sample point to obtain a final fusion prediction value of the sample point; and summarizing the final fusion prediction values of all sample points to output an aerodynamic coefficient prediction result and a corresponding interpretability analysis report.