1082 results about "Feature (machine learning)" patented technology

This application discloses a method and apparatus for constructing a multi-scale adaptive gating MambaPlus network, belonging to the field of artificial intelligence and machine learning technology. The method includes: initializing the network configuration and constructing the basic structure; preprocessing the input data to generate a standard dataset; mapping the input data to the hidden space via an input mapping layer, and extracting backbone features from the Mamba backbone; constructing at least two parallel scale branches in the hidden space to obtain a multi-scale feature set; inputting the backbone features and multi-scale features into an adaptive gating module, dynamically allocating weights and adaptively fusing them through a hierarchical gating mechanism to generate fused features; further enhancing the features through cross-scale attention and feedforward enhancement, and then superimposing the residuals to generate the final discriminative features; finally, completing category prediction and model training evaluation. This application, while retaining the advantages of Mamba's long-range dependency modeling, addresses the problems of insufficient utilization of multi-scale information, poor adaptive feature fusion, and low robustness in complex scenarios.

Systems and methods for performing a medical imaging analysis task using a foundation model are provided. One or more 3D (three-dimensional) medical images each comprising a plurality of 2D (two-dimensional) slices are received. A first set of features is extracted from the plurality of 2D slices of the one or more 3D medical images using a machine learning based encoding network. Each respective feature of the first set of features is resampled based on a spatial location of one or more pixels of the 2D slices from which the respective feature was extracted. The resampled first set of features is encoded into a second set of features. A medical imaging analysis task is performed based on the second set of features. Results of the medical imaging analysis task are output.

This information processing device comprises: a similarity acquisition unit that acquires a similarity between an explanatory sentence explaining a feature quantity in machine learning training data and a definition sentence indicating a definition of correct answer data in the training data; and a determination unit that determines, on the basis of the acquired similarity, whether or not to delete the feature quantity from the training data.

The application provides an amorphous alloy magnetocaloric performance prediction method and system based on interpretable machine learning, relates to the technical field of material science and engineering, and comprises the following steps: step 1, constructing an amorphous alloy magnetocaloric performance data set, wherein the data set comprises alloy composition data, physical and chemical descriptors, external conditions and structural characteristics of a plurality of amorphous alloy samples and corresponding experimental values of maximum magnetic entropy change and Curie temperature; step 2, screening and optimizing the features in the data set, eliminating the collinearity between the features and performing recursive feature elimination to determine the final feature subsets for predicting the maximum magnetic entropy change and the Curie temperature respectively; and step 3, establishing machine learning models for predicting the maximum magnetic entropy change and the Curie temperature respectively through the final feature subsets, and training and optimizing the models to obtain the trained prediction models. The application realizes a closed loop of high-precision performance prediction, physical mechanism analysis and directional composition design.

This invention relates to the field of machine learning technology, specifically to an intelligent analysis method, system, device, and medium for musical instrument playing techniques. The method includes the following steps: collecting standard performance audio of the target instrument and typical environmental noise; fusing them to construct a priori spectral feature benchmark library; collecting live audio and filtering it to obtain a clean performance spectrum matrix; extracting multi-dimensional features and weighting them to generate a parameter set; and matching the benchmark library data to calculate and generate performance technique discrimination coefficients. In this invention, a priori spectral feature benchmark library is constructed by fusing standard performance audio and environmental noise. Constrained whitening filtering and hierarchical feature screening remove stable and sudden noises, fully preserving effective performance spectrum features directly related to playing techniques. Fine-grained characterization of playing techniques is achieved through joint weighted quantization of time-domain and frequency-domain harmonic multi-dimensional features. Accurate discrimination of playing techniques is completed through dynamic feature matching, significantly improving the accuracy and anti-interference stability of recognition in complex scenarios.

The present application relates to the technical field of electromagnetic oven glass panel manufacturing, and discloses an electromagnetic oven glass thermal stress simulation method based on machine learning.The method comprises the following steps: deploying a sensor group on a toughened cooling production line, continuously collecting the temperature, stress, thickness, and operating parameters such as air grid wind speed, air pressure, and nozzle angle of the glass ribbon, and forming a time series data set.For each monitoring time, the preliminary correlation degree of stress and operating parameters is analyzed, and the number identification of a single glass panel and its unique stress response mode are used for calibration to accurately quantify the influence of individual differences.The non-thermal stress feature occurrence probability is introduced to screen out the data of key monitoring time points with small interference.Based on the pure data, a machine learning simulation model is constructed, the thermal stress and non-thermal noise are separated, and the accuracy of the simulation model and the guiding value for actual production are improved.

The application discloses a backdoor attack method based on color frequency injection and adaptive local enhancement, and relates to the technical field of machine learning and artificial intelligence security. The method comprises the following steps: introducing low-frequency color offset and weak high-frequency signal into an image in a CIELAB color space to perform global color-frequency injection; using a pre-trained proxy model to locate a high-sensitive perception domain of the model through mixed evaluation of gradients and class activation maps, and generating a binary mask; in an HSV color space, respectively applying nonlinear stretching factors to saturation and brightness of the sensitive domain based on the mask to perform adaptive local enhancement; using Gaussian smoothing, adaptive noise and histogram matching to eliminate edges and statistical abnormalities caused by local enhancement, completing compensation color enhancement to generate a poisoned image; and modeling a trigger core parameter as a constrained optimization problem, and using a particle swarm optimization algorithm to jointly dynamically update the trigger core parameter to obtain an optimal strategy. The application anchors the trigger feature depth in the core semantic area of the model and lurks in the normal data manifold, guarantees a high attack success rate, realizes extreme visual and feature concealment, and has strong anti-defense robustness.

Disclosed are various embodiments for analyzing machine learning models. A selection is obtained of a first tuple comprising a first feature vector and a first result generated by a machine learning model and a second tuple comprising a second feature vector and a second result generated by the machine learning model. Then, a plurality of emulated feature vectors are generated. Next, a plurality of emulated results are generated. Subsequently, a plurality of emulated decision instances are generated. Next, a decision tree is built based at least in part on the first tuple, the second tuple, and the plurality of emulated decision instances. Finally, an importance of each feature on the decision tree is computed.

There is described herein systems and methods for modifying parameters of a deep learning algorithm to improve its carbon footprint and / or water footprint and novel methods of extracting features from code and datasets relating to the deep learning algorithm. The systems and methods described estimate the carbon footprint and / or water footprint and efficiency and performance of executing the deep learning algorithm by using surrogate machine learning models. Using the estimates, the system and methods determine at least one recommendation to modify one or more parameters of the deep learning algorithm to reduce the carbon footprint and / or water footprint of the deep leaning algorithm while increasing or maintaining the efficiency and performance of the deep learning algorithm. The system provides pre-execution estimation of environmental footprint and performance, enabling environmentally informed optimization not available in existing tools.

The application relates to the technical field of artificial intelligence, and discloses a data training model parameter optimization method and system based on machine learning, which comprises the following steps: obtaining sample data of a current data training batch, and extracting statistical features of the sample data; calculating the difference degree of the statistical features and historical statistical features, wherein the historical statistical features are determined based on statistical features of a plurality of historical data training batches in a preset sliding window; judging whether to enter a new data distribution stage according to the difference degree; when it is judged that the new data distribution stage is entered, adjusting the parameter optimization strategy of the data training model, and performing parameter optimization update of the data training model based on the adjusted parameter optimization strategy. The method can monitor the change of training data distribution in real time, and actively adjust the parameter optimization strategy when a significant change is detected, so that the training process can dynamically adapt to the time-varying characteristics of data, thereby improving the stability and convergence speed of model training and obtaining better model performance.

A computerized system and method may process and predict stress levels for audio data using a machine learning based framework. A computerized system including a processor and a memory may calculate a buffer length based on a plurality of audio attributes (e.g., of a given audio input or data item), extract an audio buffer from an audio data item based on the calculated length, and predict, using a machine learning model, a stress level for the audio buffer or data item. Some embodiments of the invention may include extracting a buffer of a length determined dynamically for different audio inputs, e.g., to ensure coherency between audio attributes or features extracted from different audio inputs having different audio characteristics. In some embodiments, audio features which may be considered by the model may include, e.g., a plurality of gradients between mel-frequency cepstrum coefficients computed for relevant audio buffers or inputs.

The present application belongs to the technical field of three-dimensional fluorescence spectrum analysis, chemometrics and machine learning, and discloses a traditional Chinese medicine harvest period identification method, system and electronic equipment fusing three-dimensional fluorescence and machine learning, aiming at the problems of long time consumption of traditional Chinese medicine harvest period identification methods, insufficient precision of existing spectrum analysis and weak model generalization ability, comprising: obtaining three-dimensional fluorescence spectrum data of the traditional Chinese medicine to be identified; after background and scattering subtraction, extracting fluorescence intensity through a weight self-adaptive alternating trilinear decomposition algorithm; substituting the fluorescence intensity into an improved prediction model to output a harvest period prediction value. The present application combines the fast and efficient advantages of three-dimensional fluorescence spectrum with the precise analytical ability of high-order matrix decomposition and machine learning, realizes fast, accurate and stable identification of the harvest period of traditional Chinese medicine through feature weighting screening, batch difference correction and harvest period interval penalty optimization, adapts to batch detection requirements, reduces detection cost and provides technical support for quality control of traditional Chinese medicine.

Systems and methods for performing a medical imaging analysis task conditioned on multi-domain medical images with missing modalities are provided. 1) one or more medical images each in a different domain and 2) a domain code defining a presence of the different domains in a set of predefined domains are received. One or more weights are determined based on the domain code. One or more parameters of a machine learning based encoder are updated based on the one or more weights. Features are extracted from the one or more medical images using the machine learning based encoder with the one or more updated parameters. A medical imaging analysis task is performed based on the extracted features. Results of the medical imaging analysis task are output.

The present application provides a multi-modal deep learning network-based overlay mark asymmetry compensation method and apparatus. Manufacturing data of different modalities is acquired and then inputted into a multi-modal deep learning network for processing, so as to obtain a prediction error of overlay mark asymmetry, wherein the multi-modal deep learning network performs feature extraction on the manufacturing data of different modalities, performs feature fusion on vector features corresponding to different modalities, and associates and combines physical parameters of actual production in a manufacturing process with parameters observed by a device, so as to obtain the prediction error of overlay mark asymmetry, thereby compensating for an actual overlay mark in actual production. By introducing a multi-modal learning network, a variety of data features are fully mined and utilized, the generalization ability of a model is enhanced, and the intelligence of the device is realized by means of the machine learning technology, thereby providing an effective traceability and localization solution for process problems in wafer production.

This application discloses a method, apparatus, electronic device, and medium for predicting the temperature of wind turbine generator bearings. The method may include: identifying characteristic variables that significantly affect the temperature of the generator bearings based on raw data from the wind farm; establishing a dataset based on the characteristic variables and the generator bearing temperatures, and then establishing a random forest model to evaluate the importance of each characteristic variable to the temperature; determining optimized characteristic variables based on their importance to the temperature, and then establishing an optimized dataset based on the optimized characteristic variables and the generator bearing temperatures; establishing an LSTM network, training it using the optimized dataset, and predicting the generator bearing temperatures based on the trained LSTM network. This invention combines feature optimization and machine learning algorithms, effectively improving the accuracy of wind turbine generator bearing temperature prediction.

This invention relates to a method and apparatus for centroid localization of cultural relics by fusing three-dimensional imagery and spectral data. The method includes: acquiring three-dimensional point cloud data and spectral imagery data of the cultural relic to be located; extracting geometric features from the three-dimensional point cloud data and spectral features from the spectral imagery data, and registering and fusing the geometric and spectral features; inputting the registered and fused geometric and spectral features into a pre-trained machine learning model to obtain the geometric and spectral feature weights for each point, thereby obtaining the centroid calculation result of the cultural relic. Compared with existing technologies, this invention overcomes the limitations of a single data source in the centroid localization of complex cultural relics, significantly improving the accuracy and robustness of centroid localization.

The application provides a rock slice image data management method, device and equipment, and relates to the technical field of geological exploration. The method comprises the following steps: acquiring image data and original identification data of a rock slice to be managed, which have first annotation information; identifying key features of the rock through machine learning based on the first annotation information and an identification standard to obtain second annotation information, wherein the second annotation information contains information carrying a first annotation information label; preliminarily checking the original identification data by using the second annotation information carrying the first annotation information label to obtain a checking result and annotation deviation information; performing secondary checking on the original identification data by using the second annotation information not carrying the first annotation information label to obtain annotation missing information; and managing the image data based on the checking result, the annotation deviation information and the annotation missing information in combination with double-expert collaborative identification. The application can improve the accuracy of rock slice image data management.

Systems and methods for evaluating progression of an anatomical object over a plurality of timepoints are provided. Longitudinal medical images of an anatomical object of a patient acquired over a plurality of timepoints are received. For each respective timepoint of the plurality of timepoints, features are extracted from the longitudinal medical images acquired at the respective timepoint using a machine learning based feature extractor network and the anatomical object in the longitudinal medical images acquired at the respective timepoint is analyzed based on the extracted features using a machine learning based prediction model. Progression of the anatomical object over the plurality of timepoints is evaluated based on results of the analyses using a machine learning based progression model. The evaluation of the progression of the anatomical object over the plurality of timepoints is output.

To provide an inspection device that can more accurately inspect the appearance of an object. [Solution] An inspection apparatus according to one aspect of the present disclosure includes: an acquisition unit that acquires an image of an object; a difference calculation unit that generates a difference map based on a comparison image which is an image of the object stored in advance and the image; a learning unit that generates a learning model that outputs the degree of abnormality of the object shown in the image in response to the input of the image by performing machine learning using a plurality of the comparison images; an abnormality calculation unit that generates an abnormality map for the object using the learning model; a coupling unit that generates a combined abnormality map based on the difference map and the abnormality map; and a determination unit that determines the state of the object based on the feature portion included in the combined abnormality map.

This invention discloses an intelligent calculation method for the thermal compression ratio of a diesel engine based on machine learning, comprising the following steps: S1, acquiring diesel engine thermal operation data and generating a state data sequence; S2, performing continuous-time state encoding on the state data sequence to construct a state representation sequence; S3, performing feature contribution analysis on the state representation sequence to generate a correlated feature sequence; S4, performing state correlation propagation and compression ratio optimization calculation on the correlated feature sequence to obtain a predicted compression ratio sequence; S5, constructing a theoretical pressure curve and an in-cylinder pressure curve based on the predicted compression ratio sequence, and calculating the curve deviation sequence; S6, performing feedback updates based on the curve deviation sequence and outputting the calculation results. This invention, combined with dual-state flow liquid neural networks, possesses advantages such as strong adaptability to thermal engine states, high accuracy in compression ratio calculation, high stability in complex operating condition analysis, and strong continuous dynamic optimization capability.

A data processing method, apparatus, device and storage medium are provided. The method proposed herein includes: dividing a model input of a machine learning model into multiple segments; for a segment in the multiple segments, determining key-value replacement information of the segment, the key-value replacement information indicating a degree of association of one or more historical data units as key features with the segment, the one or more historical data units including data units located before the segment in multiple data units; obtaining a segment attention output for the segment based on at least part of the one or more historical data units according to the key-value replacement information; and obtaining a model output corresponding to the model input based on multiple segment attention outputs corresponding to the multiple segments. In this way, it is beneficial to improve the sparsity upper limit of the attention mechanism.

A system for predicting clinical trial participation may comprise a participation probability (PP) server connected to a machine learning (ML) module. The PP server may receive clinical trial (CT) parameters from a user device. The PP server may query a patient database using the CT parameters to identify potential participants. The PP server may collect social media data for the potential CT participants. The PP server may generate feature vectors based on the social media data. The PP server may provide the feature vectors to the ML module and receive participation probability index (PPI) scores for the potential CT participants from the ML module. The PP server may generate a ranked list of likely CT participants based on the PPI scores. The PP server may provide targeted content to potential participants with low PPI scores. The PP server may update the PPI scores based on engagement with the targeted content.

This invention provides a data-driven method and system for optimizing influenza vaccine transport, relating to the fields of data processing and intelligent optimization. The method includes acquiring multi-source dynamic data corresponding to the influenza vaccine transport task, and performing time alignment and feature construction on the multi-source dynamic data to obtain feature vectors representing a set of candidate transport links and a set of candidate transport nodes. The candidate transport links represent the transport modes and path combinations from the starting node to the destination node. This data-driven method and system for optimizing influenza vaccine transport establishes a data-driven machine learning prediction model by performing time alignment and feature construction on multi-source dynamic data such as road traffic, weather, transport node operational capabilities, and cold chain status. This model jointly predicts transport time, node waiting time, and temperature exceedance risks, achieving accurate characterization of the timeliness and cold chain risks in the influenza vaccine transport process.

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.

The application discloses a log analysis method, device and equipment and a readable storage medium. The method comprises the following steps: collecting logs; extracting feature words from each log to obtain feature words in each log; corresponding feature words in each log form a feature word set of each log; according to the feature word set of each log and a seed set containing multiple seeds, determining a corresponding seed of each log, and classifying each log into a class where the corresponding seed of each log is located. The technical scheme disclosed by the application extracts feature words from logs to extract key information from the logs, classifies the collected logs according to the feature word set of each log, obtains key information and important information such as the category of the logs from the collected logs, so that the operation and maintenance personnel can query and use the logs faster and better, and the result obtained by analyzing the logs can also be used as a corpus for machine learning related to automatic operation and maintenance.

The application discloses a sealant detection model incremental learning system for an industrial environment and a method thereof, and belongs to the technical field of industrial detection and machine learning. The system comprises an NG sample collection module, a knowledge distillation update module, a stable plasticity balance module, a feature invariance constraint module and a model evolution management module. Through a selective knowledge distillation strategy, the model learns new defect patterns while maintaining known defect recognition ability. Through a stable plasticity balance framework, the learning strategy is dynamically adjusted. Through feature invariance constraint, cross-batch stability is ensured. The application improves the detection rate of the detection model from 89% to 96.3%, and reduces the false positive rate by 43%.

A method for tracking downtime of a production machine (12) is provided, the method comprising the following steps: - receiving sensor data (SD) and production target data (PTD) from the production machine (12); - combining the sensor data (SD) and production target data within a certain time period (40) to provide combined data (D) and calculating feature data (CD) of the combined data (D); - determining whether the combined data (D) originates from a downtime period (44) of the production machine (12) based on the feature data (CD); and - characterizing the downtime period (44) using a machine learning module (48) implemented in a control unit (32), the machine learning module (48) setting the cause (R) of the downtime period (44) as an output value. Furthermore, a system (19) for tracking downtime of a production machine (12) is also shown.