92 results about "Proactive learning" patented technology

The present disclosure relates to systems, non-transitory computer-readable media, and methods that a implement a framework for active learning to discover pairwise interactions via representation learning. Indeed, in one or more implementations, the disclosed systems generate a first individual perturbation embedding from a first representation of a first cell exposed to a first perturbation and a second individual perturbation embedding, from a second representation of a second cell exposed to a second perturbation. For instance, the disclosed systems combine the first individual perturbation embedding and the second individual perturbation embedding to determine a predicted pairwise embedding. Moreover, in some instances, the disclosed systems generate a pairwise embedding from a representation of a cell exposed to both the first and second perturbation. Additionally, from comparing the predicted pairwise embedding with the pairwise embedding, the disclosed systems generate a measure of biological interaction of the first and second perturbation.

The application belongs to the field of semiconductor process, and discloses a high-speed microscopic design method for semiconductor process based on an artificial intelligence algorithm. The application firstly trains a high-dimensional potential energy surface model of a system by using an artificial intelligence algorithm of deep learning and active learning, then tests the accuracy of the model, and under the premise of ensuring high accuracy of the first principle calculation, uses the potential energy surface model to optimize the semiconductor process. The model based on deep potential can more quickly calculate phase transition of atomic lattices, diffusion of impurity atoms, ion implantation, photolithography, etching, thin film deposition and growth problems in the semiconductor process. The application realizes fitting of the high-dimensional potential energy surface based on the algorithm of artificial intelligence, can perform microscopic design on the semiconductor process, and has the characteristics of high calculation accuracy, fast calculation speed, short design period and low design cost.

The application discloses an electronic medical record and a cataloging and classifying method, system and device thereof. The method comprises the following steps: obtaining standardized data by preprocessing multi-format medical record original data through a medical OCR model, a VAE anomaly detection algorithm, a medical knowledge graph word segmentation tool and a Transformer term standardization model; checking semantic rationality, extracting multi-dimensional features, and obtaining a comprehensive feature vector after strengthening and fusing; constructing a transfer learning classification model and training the model, inputting the feature vector to generate cataloging information; and finally, evaluating the classification result through an active learning mechanism, combining artificial labeling data and a causal forest dynamic updating framework, and incrementally learning and optimizing the model performance. The application solves the problems of low multi-format data extraction accuracy, insufficient term standardization and poor model generalization in the prior art.

The application discloses a dynamic reliability evaluation method for a permanent magnet planetary gear transmission system, and belongs to the technical field of reliability analysis of electromechanical coupling transmission systems. The method is characterized in that: firstly, the service process of the system is divided into stages, and a random variable model is established; based on the wear evolution relationship, parameters of each stage are obtained through parameter mapping; then, stage high-fidelity unified electromechanical dynamics solving is performed, and frequency-related characteristic indexes are extracted; a multi-failure mode limit state vector is constructed, and a stage failure event is defined; on this basis, a multi-response Gaussian process proxy model is used for fitting and iterative updating of the limit state function through stage-by-stage active learning; the U function is used for initial screening of candidate samples, EFF is calculated and sorted, and a new sample point is selected according to the Top-K criterion to update the proxy model, until the convergence criterion is met; finally, the dynamic failure probability of the system is calculated by using the updated proxy model. The application realizes efficient and accurate evaluation of the failure probability of multiple failure modes, reduces the calculation cost, and improves the engineering applicability.

Presented is an invention in which visible light is substituted for damaging radiation, enabling students to learn about radiology through active learning (e.g. by manipulation of a study's subject and seeing, in real time, the resulting changes in the generated image) but without exposure to damaging radiation. The invention also allows student and senior practitioners to practice radiological techniques (e.g. patient positioning, fluoroscopy) without radiation exposure. The invention also allows demonstration of the relationship between a three-dimensional object and its two-dimensional depictions, such as between a machine part and that part's views on a technical drawing.

This invention discloses a near-field scanning detection method based on active learning to obtain conducted coupling paths. The method includes: constructing a near-field scanning detection platform; placing the circuit board under test (PCB) in a microwave anechoic chamber; and using the near-field scanning detection platform directly above the PCB to perform detection; obtaining the detection results using an active learning-based near-field scanning method; and obtaining a near-field current distribution map on the PCB at continuous time intervals based on the detection results, thereby realizing the detection of conducted coupling paths. This invention utilizes active machine learning methods to achieve more efficient near-field scanning detection, concentrating scanning points in areas with strong currents and accurately reconstructing the actual current changes of the PCB using the scan results of sparse points, greatly improving detection efficiency. By using near-field scanning detection to reconstruct the current changes on the circuit board, the coupling paths of conducted interference currents on the PCB can be quickly analyzed, meeting the requirements for detecting electromagnetic interference in systems in practical engineering.

The application discloses a high-entropy alloy target performance reverse rapid proportioning method based on active learning, belongs to the technical field of material genetic engineering and artificial intelligence, and comprises the following steps: establishing component normalization, preparation process and phase stability engineering constraints; constructing digital characterization and continuous design space of high-entropy alloy components; constructing and pre-training a differentiable performance predictor based on an attention mechanism; fixing the predictor parameters, constructing a composite loss function driven by the target performance, iteratively optimizing the component vector through gradient back propagation and normalized projection, and outputting the theoretical optimal candidate component; screening high-value samples based on an active learning strategy to verify and incrementally update the model, and forming a closed-loop optimization. The application breaks through the limitations of traditional forward prediction models, realizes rapid and accurate reverse design from target performance to component proportioning, and effectively solves the combination explosion and data scarcity problems in the high-dimensional design space of high-entropy alloys.

The application discloses a kind of based on active learning and man-machine collaborative optimization's massaging robot control method, it is related to artificial intelligence, medical rehabilitation robot and complex contact type operation control cross technical field, this method relies on multimodal perception system, core computing hub and remote man-machine collaborative interface control platform, in turn complete the initial massaging strategy model construction of multi-source heterogeneous expert data, the uncertainty measurement based on bayesian neural network, man-machine collaborative key frame active learning trigger, reinforcement learning reward function design of fusion patient biological feedback and incremental learning of impedance control parameter and force-position hybrid output;The application greatly improves the clinical safety of massaging robot, exponentially reduces data acquisition cost, realizes individual physiological closed-loop rigid-flexible massaging, also overcome the catastrophic forgetting problem of neural network, endow system lifelong learning ability.

This invention relates to an active learning-based data-free black-box attack method and system based on multidimensional value assessment, belonging to the field of artificial intelligence security technology. This method constructs a pre-emptive "sample screening funnel," utilizing a local substitution model to perform multidimensional assessments of sample boundary approximation, information uncertainty, and geometric diversity before sending images to a commercial cloud API. Only high-value samples are selected for querying, thereby achieving low-cost, high-efficiency model theft and adversarial attacks. This invention ensures the diversity and training stability of data-free generated samples, significantly improves the transfer success rate of adversarial examples, and achieves "low-cost, low-risk" economical attacks. It has strong versatility and can be seamlessly integrated into various existing data-free attack frameworks, facilitating deployment and implementation in practical security assessment systems.

The application discloses a skin type prediction method for small sample data autonomous learning, and relates to the technical fields of deep learning and medical beauty skin type. The initial model is obtained by a convolutional neural network through a small number of labeled samples, and the subsequent new samples are accepted through the initial model, and whether the new samples have labeling value is judged according to the prediction result of the model, so that only the samples determined to be inaccurate by the model are selected for labeling and retraining. Compared with the existing skin type method, a single model can be used to obtain a more perfect result, the model is continuously iteratively optimized by using the enhanced unit in the actual use process, and the idea of active learning is used to select the samples with greater uncertainty for labeling, so that the number of samples needing expert labeling can be greatly reduced, and the labor cost is reduced. According to the reliability module, when the samples are expanded, the skin type with insufficient reliability is preferred, so that the final typing effect is improved.

The application provides a high-performance water-based acrylic emulsion formula screening method, comprising the following steps: constructing a multi-source heterogeneous iterable database and differentiating storage; extracting key features from the database and performing cross-scale feature conversion and screening to obtain important key features; taking the important key features as input and the target performance as output, establishing a machine learning model and optimizing; using the optimized model to perform high-throughput virtual screening on virtual formula and process combinations, screening out candidate formulas with synergistic optimization of flame retardancy, hardness and adhesion, preparing water-based acrylic emulsion samples, obtaining the flame retardancy, hardness and adhesion data of the samples, completing verification and analysis; according to the verification data, further updating the database and model based on an active learning strategy, and performing explainability analysis of the machine learning model to guide subsequent experiments; repeating the above steps until the target high-performance water-based acrylic emulsion formula is obtained.

Embodiments of the present application provide a model training method and device, electronic equipment and storage medium, and relate to the technical field of computers. The method comprises: retrieving a set of documents from a preset document library based on a set of keywords; clustering the set of documents to obtain a plurality of subsets of documents; extracting at least one document from each subset of documents as a first document; determining positive samples and negative samples from the first document through a plurality of large language models and training an initial model to obtain a target model. After training the baseline model on the few-sample labeled dataset of the multi-model voting, the baseline model is enhanced using an active learning strategy. First, the baseline model is used to evaluate a randomly selected set of documents, and the sample with the lowest prediction is iteratively annotated. Using the large model debate idea, when the accuracy, recall rate and other indicators of the model no longer have significant improvement, the iteration is stopped, so that the trained model has high target field recognition ability.

The application provides a first-person visual angle auxiliary perception method for people with visual field defects, and relates to the technical field of visual technology.The application realizes automatic recording and auxiliary perception of target objects based on initial calibration and recording habits of users, and comprises the following steps: initialization calibration, user habit learning, multi-modal data acquisition, automatic recording and auxiliary perception.The method combines personalized needs of users and characteristics of visual field defects, and improves the perception efficiency and integrity of target objects for people with visual field defects through active learning and automatic recording.

This invention discloses an active learning method for semantic segmentation of remote sensing images of road cracks based on a self-supervised dual representation mechanism, belonging to the field of computer vision and intelligent interpretation technology of remote sensing images. The method includes: extracting global semantic distribution features of unlabeled sample pool images using a first self-supervised model with frozen parameters; calculating pixel-level reconstruction error maps using a second self-supervised model with frozen parameters; generating a comprehensive sampling value score by weighted fusion of the two types of self-supervised features, with a cold-start hybrid sampling strategy adopted in the first iteration and dynamic adjustment of fusion weights in subsequent iterations; selecting a high-value sample set from the unlabeled sample pool and performing pixel-level annotation; and independently training the semantic segmentation main network using the labeled sample set, with the self-supervised model not participating in gradient updates throughout the process. This invention achieves complete decoupling between the sampling module and the segmentation model, significantly improving crack edge segmentation accuracy and model convergence efficiency with low annotation costs.

This invention discloses a data annotation method and system for autonomous driving, relating to the field of data annotation. First, multi-source sensor data is acquired and initially annotated using an automated model. Then, a joint optimization algorithm decomposes and reconstructs features, improves boundary annotation accuracy, and identifies low-confidence regions. Based on a deep active learning strategy, prediction entropy, Bayesian divergence, and task-level uncertainty are fused to screen high-value samples. Ground truth labels are obtained through manual verification, while low-confidence regions are optimized to generate supplementary labels. These two types of labels are used as incremental training data, and the model mapping matrix is ​​updated through topological residual projection. Finally, the model is deployed for road testing, and problematic data is collected, triggering a new annotation optimization process to form a closed-loop iteration. This invention improves annotation accuracy and efficiency, achieves efficient incremental model updates, and constructs a continuously evolving annotation closed loop, providing support for the iteration of autonomous driving models.

The application discloses a kind of power inspection image analysis algorithm based on deep learning, belong to computer vision and artificial intelligence technical field.The method includes: constructing hierarchical annotation dataset;Design and train global-to-local double-flow feature fusion network, which dynamically weights and fuses global context and local detail features through adaptive feature fusion module;Multi-task collaborative learning is carried out, and the model is optimized in combination with self-supervised pre-training;Based on power equipment knowledge graph, the logical reasoning and verification of visual recognition result are carried out;A dynamic updating closed loop is established by fusing active learning and elastic weight consolidation technology, so that the system continuously evolves.The system should include data management, model training, image analysis, knowledge reasoning and active learning modules.The application combines the perception, reasoning and evolution ability in depth, significantly improves the accuracy, robustness and explainability of power equipment defect identification, and has the ability of continuous self-optimization from practical application data.

This invention discloses a multi-dimensional privacy information perception and collection method. It constructs a formal description model of multi-dimensional privacy information, characterizing privacy information from multiple perspectives, including its composition, attributes, permissions, and control. Based on active learning and transfer learning algorithms, it builds a metadata intelligent perception, identification, and labeling method adaptable to various application scenarios and privacy protection strategies. This patent's advantages in adapting to multiple scenarios and information types allow it to adapt to the impact of changing application scenarios on privacy properties, reduce the concealment of privacy information caused by multiple information types, and improve the accuracy of privacy information perception. It focuses on perceiving multiple types of privacy information, comprehensively perceiving different privacy types such as text, images, and sound. Compared to existing single-type privacy perception methods, it is more beneficial for subsequent privacy strategy formulation and privacy control implementation, improving user privacy security. It has strong privacy perception capabilities, can process different types of data in parallel, and offers accurate privacy positioning and fast data processing.

The application discloses a kind of server mainboard memory stick plug test system, including active acoustic scanning unit, for emitting sound wave and gathering reflected sound wave;Adaptive noise suppression unit, for noise suppression to original reflected sound wave;Acoustic fingerprint extraction unit, for extracting acoustic feature vector;Meta-learning adaptation unit, for generating personalized encoder based on global meta-learner parameters and heuristic acoustic feature vector;Health state classification unit, for outputing shell health grade;Uncertainty processing unit, for evaluating confidence and triggering active learning;Privacy protection federal learning unit, for updating model after adding noise in local update collaborates with central server.The application realizes non-contact high-precision detection, zero-sample rapid adaptation, cross-vendor privacy protection data collaboration, active learning lossless verification and multi-objective optimization scheduling, significantly improves detection accuracy, change type efficiency and production line flexibility, with high industrial application value.

The present application provides a shale in-situ conversion property full cycle evolution identification and characterization method, which belongs to the technical field of shale in-situ conversion. The present application determines the basic parameters by selecting shale samples, heats at a preset heating rate under preset confining pressure hole pressure and fluid boundary conditions, and obtains dynamic evolution parameters by online monitoring and staged sampling, identifies the evolution stage, calculates the candidate boundary confidence according to the evidence fusion theory, combines with the preset stage semantic rules, uses the hierarchical priority strategy to resolve conflicts to generate stage division results, uses the active learning strategy combined with Gaussian process regression to quantify uncertainty and encrypt experimental temperature points, trains the physical information constraint artificial intelligence prediction model, and outputs the property dynamic evolution curve, which solves the technical problem that the property evolution stage boundary in the shale in-situ conversion process is difficult to accurately identify.

The application discloses a stamping process intelligent decision-making method of man-machine cooperation, constructs a stamping forming prediction model based on a physical information neural network, embeds Hill48 yield criterion, Swift hardening model and friction law into a loss function in the form of a regular term, trains process sensitivity gradient by using small sample data and outputs the process sensitivity gradient, calculates information gain by using an expectation improvement or a confidence upper bound acquisition function and recommends a die trial point based on model cognitive uncertainty, triggers manual confirmation when uncertainty is higher than a threshold value, executes process parameters, collects a stamping force-displacement curve in an initial die trial stage, inverses sheet metal parameters by using an extended Kalman filter to correct model input, records results and feeds back and updates the model, solves a small sample learning problem by embedding physical constraints, reduces die trial times by active learning, realizes material self-adaptation by online inversion, and forms a closed-loop optimization of prediction-die trial-inversion-updating.

The application provides a dialog virtual assistant tuning method and system, and relates to the technical field of artificial intelligence and natural language processing, and the method comprises the following steps: determining a target network structure from a preset neural network set according to feature information of training corpus of the dialog virtual assistant; performing hyperparameter optimization and model training according to the training corpus to obtain a target model under a target hyperparameter combination; extracting a target sample set to be labeled from a log database; incorporating the target sample set labeled by manual labeling into a training data set, and performing model incremental updating of the target model according to the training data set and the target hyperparameter combination. The application improves the universality through adaptive selection of the network structure, improves the tuning efficiency through automatic hyperparameter tuning, guarantees the model performance, forms a complete closed loop of data, model, feedback and updating through the active learning mechanism, realizes the continuous evolution of the model, and thus improves the prediction accuracy of the model.

The application belongs to the field of structural reliability analysis and uncertainty quantification, and specifically discloses an active learning reliability analysis method based on spectral normalized Gaussian process, which comprises the following steps: generating a Monte Carlo sample set according to a joint probability density function of random variables; then constructing an initial training data set and a spectral normalized Gaussian process (SNGP) surrogate model of a structural performance function; predicting the Monte Carlo sample set by using the surrogate model, estimating a system failure probability and screening a candidate sample pool; judging whether the model converges and whether the system failure probability meets the accuracy requirement according to the prediction result; if yes, outputting the surrogate model and the system failure probability estimation result, and completing the analysis. The application can significantly reduce the number of calls to the original calculation model in complex high-dimensional nonlinear reliability analysis, greatly improve the analysis efficiency while ensuring the accuracy, and is suitable for reliability analysis and design optimization of high-dimensional strong nonlinear systems in the fields of aerospace, vehicle engineering and the like.