3 results about "Feature selector" patented technology

This invention discloses a rapid method for detecting and analyzing the composition of geological experimental samples, belonging to the field of geological analysis and testing technology. The method includes: acquiring multimodal data such as laser-induced breakdown spectroscopy, X-ray fluorescence spectroscopy, and Raman spectroscopy; performing baseline correction, noise filtering, and feature extraction through multi-agent collaborative processing; constructing a deep reinforcement learning feature selector to dynamically screen cross-modal shared sparse feature sets and heterogeneous complementary feature sets; using an improved ant colony algorithm and a Bayesian neural network for adaptive weighted fusion; constructing a physical information constraint network embedded with geochemical constraints to generate preliminary predicted values; and finally correcting abnormal predicted values ​​using a graph neural network topology corrector to output the final compositional analysis results. This invention integrates multi-agent algorithms with prior physical knowledge, significantly improving the accuracy, robustness, and interpretability of geological sample compositional analysis, and solving the problems of low analytical accuracy, poor generalization ability, and lack of physical consistency in existing technologies.

This invention relates to the field of defect detection technology, specifically to a PCB defect detection method for industrial incremental scenarios. It includes a feature fusion method based on a lightweight segmentation strategy that combines random pruning with separate processing of strong and weak features. This strategy simplifies redundant computation by using a feature selector, and differentiates and recombines defect features of varying saliency during the feature fusion stage. This significantly improves inference speed while maintaining the precision of segmentation, solving the problems of high computational cost and difficulty in capturing minute defects in existing methods. The method also includes an incremental learning approach employing a background classifier adaptation mechanism and local semantic distillation. Class-specific regularization and spatially weighted logical alignment distillation work synergistically. By dynamically calibrating the background prediction logic and constructing a pixel-level semantic relevance matrix, it achieves deep alignment between new and old knowledge and background distribution. This effectively solves the catastrophic forgetting problem caused by the evolution of PCB background texture in existing incremental learning methods, significantly enhancing detection stability and the adaptability of the enhanced model.

This application relates to a motion capture method, device, equipment, and medium for ice hockey passing and control training. The method includes: acquiring raw motion data from an inertial measurement unit (IMU), synchronizing and cleaning it to obtain a synchronization data matrix; constructing an initial spatiotemporal graph using IMUs as nodes and spatial relationships as edges; generating weight information using a LASSO feature selector, and weighting it to obtain an updated spatiotemporal graph; inputting the updated spatiotemporal graph and weights into a graph neural network for synchronous inference of motion segmentation, category, and quality assessment information; decoding the motion boundaries and category labels using the Viterbi algorithm; inputting the data into a scoring regression model to calculate a technical quality score, and outputting the motion analysis results. This method can accurately capture ice hockey passing and control motions, achieving integrated motion segmentation, classification, and quality assessment, thus improving training analysis efficiency.