14 results about "Subspace model" patented technology

The invention discloses an environmental noise classification method based on adaptive joint parameter space optimization, and the method comprises the steps: collecting a plurality of noise signals, enabling each type of signals to correspond to a specific environmental noise type, and constructing a data set; defining a joint parameter space comprising a plurality of optimization variables, wherein the joint parameter space comprises a data enhancement parameter subspace, a model network parameter subspace and a model training hyper-parameter subspace; according to training data characteristics, environmental noise classification task complexity and model deployment constraint, adaptively calculating each parameter range space; and constructing an objective function, and searching a multi-parameter optimal collaborative combination by using Bayesian optimization. According to the method, a joint parameter space is constructed, and Bayesian optimization is utilized to adaptively search a multi-parameter optimal collaborative combination of a data enhancement parameter, a model network parameter and a training hyper-parameter in a multi-model training process, so that synchronous dynamic optimization of data enhancement, a model structure and model training is realized; and finally, the performance of the neural network model in environmental noise classification is improved.

This invention discloses an image classification method, system, and medium integrating a dynamic task subspace. The method includes: acquiring an image to be classified; inputting the image to be classified into an image dynamic task subspace model to obtain an image classification result; wherein the model is obtained through the following steps: constructing a shallow image feature extractor; constructing a deep image feature extractor; constructing an image task subspace feature extractor; acquiring training images; obtaining a first super feature based on the training images and the image task subspace feature extractor; obtaining a second super feature based on the training images and the image task subspace feature extractor; calculating the classification loss using cross-entropy; calculating the knowledge distillation loss; calculating the total loss; and updating the parameters of the image task subspace feature extractor based on the total loss to obtain the image dynamic task subspace model. This invention improves classification accuracy and reduces parameter storage costs. This invention can be widely applied in the field of image recognition technology.

The invention relates to a dual three-phase permanent magnet synchronous motor model-free prediction repetitive control method and system based on a double-subspace virtual vector, and the method comprises the steps: decomposing the vector of a motor into a fundamental wave subspace and a harmonic wave subspace which are orthogonal to each other based on an obtained operation parameter, and constructing an independent super-local model in the two subspaces; constructing a linear expansion state observer in the fundamental wave subspace, and constructing a repetitive expansion state observer based on repetitive control in the harmonic wave subspace; respectively synthesizing an independent virtual voltage vector set and a decoupling virtual voltage vector set in the two subspaces; and calculating reference voltage vectors of the two subspaces, selecting an optimal virtual voltage vector and an optimal decoupling virtual voltage vector, respectively calculating optimal duty ratios, synthesizing control signals applied to each bridge arm of the inverter, and driving the dual three-phase permanent magnet synchronous motor. According to the method, model-free control of double sub-spaces is realized, model parameters of a motor system of an algorithm are controlled, and better robustness is shown when the parameters are mismatched.

The present invention belongs to the field of brain-computer interface technology application technology, and discloses a brain-computer information fusion classification method and system for shared subspace learning, wherein the brain-computer information fusion classification method includes a training phase and an inference phase; wherein the training phase utilizes paired images and brain response data, optimizes the shared subspace model parameters of images and brain responses through a comparative learning strategy of positive and negative sample sampling, and trains an image classifier; the inference phase extracts image features for classification, and achieves the application goal of the entire brain-computer information fusion classification system. The brain-computer information fusion classification system for shared subspace learning of the present invention can train shared subspaces end-to-end, achieve efficient transfer of brain cognitive information, and improve the performance of image classification tasks in complex open scenarios; through the application of "brain out of the loop", it improves efficiency and stability in real-world applications, and has broad application prospects under the new paradigm of brain-computer information collaboration.

The invention discloses a self-adaptive subspace projection suppression method based on optical pump magnetometer external noise, and belongs to the technical field of neural magnetic imaging (MEG) signal processing.The method comprises the steps that brain magnetic signals are collected through a wearable optical pump magnetometer array and preprocessed, and a signal and noise subspace decomposition model is established; integrating a model-driven noise modeling method and a data-driven noise modeling method, and constructing a noise subspace base vector; constructing a signal subspace basis vector in combination with a guide field matrix and event correlation analysis; identifying residual noise components based on event correlation analysis, and perfecting noise subspace modeling; signal noise separation is realized by adopting a hierarchical projection strategy; and analyzing the time correlation of internal and external space signals, and constructing a time projection matrix to remove residual interference to obtain a clean MEG signal. According to the invention, various types of external noise interferences can be effectively suppressed, the signal-to-noise ratio and the reliability of brain magnetic signals are remarkably improved, and the method is particularly suitable for practical application of a wearable optical pump magnetometer system.

The invention relates to a skull dissection method, a skull dissection device, skull dissection equipment and a medium. The skull dissection method comprises the following steps: S1, extracting brain tissue space prior information from a brain magnetic resonance image public data set based on a subspace model; s2, pre-segmenting the brain magnetic resonance image to be segmented based on a multi-resolution multi-map method; s3, constructing a multi-resolution position correlation network, and guiding position correlation network training based on the subspace prior in the step S1 and the pre-segmentation result in the step S2 to obtain a segmentation result; and S4, constructing a fusion network, taking the subspace priori obtained in the step S1 and the segmentation result of the multi-resolution position correlation network obtained in the step S3 as input, and fusing the subspace priori and the segmentation result to output a final skull dissection result. Compared with the prior art, the method has the advantages that the boundary definition accuracy is improved, and the skull stripping result accuracy is high.

This invention discloses a method and system for suppressing false rejection behavior and selectively forgetting harmful semantics in large language models across vertical domains based on feature subspace decoupling, belonging to the field of fine-tuning and alignment technology for large artificial intelligence models. The method includes steps for target definition and data construction, feature subspace localization, key level optimization, orthogonal decoupling weight correction, constrained norm renormalization, selective fine-tuning, false rejection calibration, and closed-loop verification monitoring and document output. The system includes modules for concept discovery and data construction, representation subspace modeling, false rejection localization and calibration, and verifiable evaluation and continuous monitoring. This invention can significantly reduce the false rejection rate and maintain a safety baseline while preserving the model's generalizability, meeting the compliance requirements of generative artificial intelligence services.

This invention discloses a multi-view, multi-label learning method based on deep feature map fusion. Addressing the limitation of single shared subspace models in fully describing all semantic information of multi-view data, this method proposes a multi-view, multi-label classification method based on deep feature map fusion. By mining the complementary relationships of instance features across multiple views and the structured symbiotic relationships of label features, it constructs a more representative instance-label structured vector representation and classifies the data by averaging the "instance-label" affinity matching results of individual views. This method enhances the structured semantic representation capability of each individual view by constructing a unified feature map structure across multiple views, fusing nearest neighbor relationships within a single view and alignment relationships across views. It emphasizes the contribution of individual views to specific semantic representations while integrating the consensus and complementary relationships of multi-view data. This effectively improves the semantic representation capability of multi-view data and has strong application value for practical data analysis and decision-making.

The invention relates to a brain magnetic resonance image segmentation method and device, equipment and a medium. The method comprises the following steps: S1, extracting brain tissue space prior information from a brain magnetic resonance image public data set based on a subspace model; s2, performing iterative segmentation on the brain magnetic resonance image to be segmented based on the brain tissue space prior information; s3, constructing a position correlation network, and based on the subspace prior obtained in the step S1 and the segmentation result obtained in the step S2, using the subspace prior obtained in the step S1 and the segmentation result obtained in the step S2 as prior information to guide the position correlation network to learn probability distribution of brain tissues; and S4, constructing a fusion network, taking the subspace priori obtained in the step S1, the segmentation result obtained in the step S2 and the segmentation result of the position correlation network obtained in the step S3 as input, and fusing the subspace priori, the segmentation result obtained in the step S2 and the segmentation result obtained in the step S3 to output a final segmentation result. Compared with the prior art, the method has the advantages of accurate segmentation result, good robustness and the like.

The invention provides a pumped storage global optimization control method and device and electronic equipment, and the method comprises the steps: building an operation mode perception and reconstruction mechanism through information-physical fusion modeling, and generating a structured scene set which can comprehensively represent the uncertainty of a power system; constructing a global optimization control model in combination with the structured scene set, and obtaining an optimal pumping and storage planning scheme of the power system through the global optimization control model; the global optimization control model adopts a main-subspace collaborative optimization architecture and comprises a main space model and a subspace model. According to the invention, the method can effectively cope with the operation constraint of a complex power system, guarantees the optimality and solving efficiency of a planning result, and solves a problem that the planning control effect of the power system is not good in the prior art.

Methods, systems, and devices for modeling a relationship between glucose sensitivity and a sensor electrical property are described herein. More particularly, the methods, systems, and devices describe partitioning an input signal feature space relating glucose sensitivity and a sensor electrical property into subspaces and training a model for each subspace. For example, the subspace models may form a mosaic of models, for which the output is more accurate than a single model.

The invention discloses an intelligent smoke-proof air curtain parameter optimization system and method based on a proxy model and reinforcement learning. The system comprises a high-fidelity simulation and order reduction module, a proxy model construction module, a reinforcement learning control strategy training module, a data acquisition module, a control execution module and an air curtain. The method comprises the following steps: establishing a multi-physics-field coupled fire smoke diffusion numerical model based on finite element simulation software; a low-order subspace model capable of being used for rapidly reconstructing system behaviors is constructed based on a reduced-order algorithm, and a performance expression data set under different control parameters is predicted; based on the low-order subspace model data, a deep learning method is driven to construct a reduced-order proxy model; based on reinforcement learning, training an optimal control strategy of the air curtain in the constructed reduced-order proxy model environment; the air speed and the spraying angle of the air curtain are optimized in real time through the control execution module, and effective control and dynamic adjustment of fire smoke are achieved. According to the invention, the smoke control efficiency, the energy efficiency level and the cross-scene adaptive capability of the system are improved.

The invention discloses a VPN handshake anomaly identification method based on self-supervised learning. The method comprises the following steps: constructing a handshake feature vector sequence from unlabeled handshake traffic; generating a protocol semantic graph and a semantic template correlation matrix based on the handshake feature vector sequence; constructing a multi-view handshake sample set according to the session, the certificate cluster and the server fingerprint; establishing a handshake coding network and normal and abnormal branch projection head self-supervision structures, and dividing training samples; in a normal branch, using improved Barlow Twins loss to learn normal subspace representation; sparse correlation, semantic decoupling and cross-branch orthogonal constraint are applied to the abnormal branches to obtain abnormal subspace representation; and updating the semantic template correlation matrix in online operation, generating an exception score for a handshake to be detected, and outputting an exception type. According to the method, the semantic template and the double-subspace model are established under the non-labeling condition, and the method is used for identifying degradation, playback and certificate replacement anomalies.

The application relates to the technical field of nonlinear dynamic system fault diagnosis, in particular to a nonlinear dynamic system fault diagnosis method and device based on a mask attention subspace model. The method comprises the following steps: a subspace coding-decoding model is constructed, wherein the encoder serves as a state estimator, and the decoder serves as an output estimator; a mask attention layer is introduced into the encoder, and the attention range is constrained through an inter-neighbor mask module; a trained MAS-Net model is used to calculate residual signals and T2 statistics of test samples; a fault detection threshold is determined based on kernel density estimation, so that fault detection is realized; and fault variable identification and isolation are realized by analyzing the differences in attention weight under normal and fault states. The application solves the problems that traditional subspace identification methods are difficult to extract long-distance dependent features and have poor fault variable interpretability, can effectively process high-dimensional, nonlinear and strongly coupled industrial data, and simultaneously realizes high-precision fault detection and interpretable fault isolation.