93 results about "Active learning (machine learning)" patented technology

The invention belongs to the field of natural language processing, and provides a pre-training large model physical examination text labeling method and system based on active learning. The method is realized through the following process: collecting and sorting structured and unstructured physical examination data, manually pre-labeling a small number of samples by an expert and standardizing a format; selecting a pre-training large model adaptive to the medical field task; designing a domain-specific cue word template to strengthen semantic understanding of the model; performing parameter optimization by adopting supervised fine tuning and low-rank adaptation technologies; an active learning closed loop is constructed, uncertain samples are selected through multiple rounds of iterative training to supplement labels, and the model is dynamically optimized by using a verification set; and the optimized model is applied to intelligently label the remaining data. According to the scheme, the key problems of low efficiency, poor rule method generalization ability, high machine learning model computing power demand and insufficient large model field adaptation of traditional manual labeling are solved, and the automatic extraction precision of medical entities such as pathological features and clinical indexes in the physical examination report can be effectively improved.

Using active learning to detect Protected Health Information (“PHI”) in documents stored as unannotated natural language data by selecting an initial chunk of text from the documents; forming a gold standard data via annotating the text by a human, the annotating identifies and tags PHI required to de-identify the text; training, using machine learning and the text before and after the annotating, a model having rules for PHI detection; querying, using a strategy, the documents to select a next chunk of text; machine annotating the text using the trained model; updating the gold standard data via correcting the machine annotation of the text by the human, wherein an amount of corrections in the updated gold standard data indicates a quality of the machine annotation; and iterating the steps starting at training, until the quality of the machine annotation is higher than a predetermined quality threshold.

The invention relates to the technical field of abnormal domain name detection, in particular to an abnormal domain name detection system based on machine learning, which comprises a data processing and feature extraction module, a model training and updating module, a real-time detection and classification module, a family clustering and analysis module and an active defense and response module. Compared with a traditional abnormal domain name detection system which usually depends on a small number of characteristics such as character frequency and domain name length for detection and is difficult to comprehensively capture abnormal characteristics of domain names, the abnormal domain name detection system has the advantages that multi-modal characteristics such as character statistics, semantic characteristics and context behavior characteristics are extracted; in addition, the weight of each feature is dynamically adjusted through an attention mechanism, the problem that a single feature is easy to avoid is solved, high-uncertainty samples are screened through an active learning selector to update a training set, the problem of sample imbalance is solved, and the generalization ability of the model is improved.

Examples may provide a method of selecting data as part of an active learning process. The method includes extracting features from different classes of medical image datasets using a self-supervised contrastive learning technique to produce a set of extracted image features. The method also includes assigning clusters to achieve label diversity using the set of extracted image features to produce a set of clusters. In addition, the method also includes selecting hard-to-contrast data points from dataset maps, wherein a hard-to-contrast data point in a given cluster from the set of clusters is harder to contrast itself with other data points in the given cluster. Examples may also provide a method of segmenting an image using a contrastive language-image pre-training (CLIP)-driven universal model. Additional methods as well as related systems and computer readable media are also provided.

Embodiments relate to system and methods involving use of a technique for managing a database for producing a material composition having a thermodynamic phase. The technique can include: receiving a binary phase diagram for each material to be used as a component of a high-entropy alloy (HEA); using one or more active learning machine learning techniques for generating a feature, the feature including: a primary feature that is representative of a probability that an HEA will exhibit a solid solution phase and / or an intermetallic phase, and a physics-based feature that is representative of a factor related to formation of a desired intermetallic HEA phase; encoding the primary feature and the physics-based feature; generating an output representation of a HEA alloy composition and phase of a predicted materials composition; and selecting a HEA composition and phase that will meet a material design criterion.

Systems and methods for training a machine learning model to classify defects with utility-function-based active learning are described. In one embodiment, one or more non-transitory, machine-readable mediums are configured to cause a processor to at least determine a utility function value for unclassified measurement images, based on a machine learning model, wherein the machine learning model is trained using a pool of labeled measurement images. Based on a determination that the utility function value for a given unclassified measurement image is less than a threshold value, the unclassified measurement image is output for classification without the use of the machine learning model. The unclassified measurement images classified via the classification without the use of the machine learning model are added to the pool of labeled measurement images. The machine learning model is trained based on the measurement images classified via the classification without the use of the machine learning model.

The invention discloses a multi-component catalyst active site prediction system and method fused with quantum embedding, and relates to the technical field of catalysis and material informatics, and the system comprises a structure and site enumeration module which generates candidate sites; the adaptive quantum embedding calculation module obtains key reaction microcosmic parameters; the unified site fingerprint and feature engineering module constructs and fuses standard site fingerprints; the physical consistency machine learning module predicts adsorption energy and other parameters and uncertainty thereof; the active learning and sample selection module selects a high-value sample optimization model; the microdynamics evaluation module calculates index values such as activity; and the multi-objective optimization and sorting module generates an optimization sorting list. According to the method, the unification of calculation precision and efficiency is realized, the problem of non-unification of locus characterization is solved, the model interpretability and extrapolation reliability are improved, and the comprehensive evaluation and optimization sorting of multi-target performance are completed.

The invention discloses a microemulsion interfacial tension efficient prediction method and system based on active learning and molecular dynamics. According to the method, 217 molecular descriptors corresponding to each molecular structure are calculated by adopting an RDKit software package, and the descriptors are used for representing molecular structure characteristics and serve as input variables of a machine learning model, so that key structure information including molecular branching degree, polarity and the like is transmitted. For an oil-water-surfactant ternary interface system, the oil-water interfacial tension in the presence of a surfactant is simulated and calculated through molecular dynamics, and an IFT value is set as a model prediction target. An active learning mechanism is introduced, and iterative sample labeling in the molecular dynamics simulation process is guided; and integrating the obtained IFT data with the molecular descriptor features, constructing a machine learning data set, and training a random forest model. According to the method, the problem of screening a high-performance surfactant layer by a middle-phase microemulsion system can be solved, and the ultra-low oil-water interfacial tension can be rapidly and efficiently screened.

Methods and systems are described herein for minimizing resource expenditure during model training using user-defined constraints in sample selection. A system may obtain user-defined target parameter values for data labeling, a user input indicative of a value added per unit of model performance improvement, and a dataset (e.g., unlabeled samples). The system may select a first subset of the dataset and may transmit a request for labeling the samples. The system may receive a first training dataset comprising label data and the samples of the first subset. The system may train a machine learning model using the first training dataset and generate a margin curve. Based on the margin curve, the system may determine whether an amount of resource usage exceeds value added and responsive to determining that it does not exceed the amount of resource usage, select a second subset of the dataset.

Various embodiments of the present disclosure provide methods, apparatus, systems, computing devices, computing entities, and / or the like for labeling data by (i) generating, using a natural language machine learning model, a labeled dataset from unlabeled data, (ii) training one or more instances of a classification machine learning model based on the labeled dataset, (iii) generating, using the one or more instances of the classification machine learning model, a plurality of validation classifications, and (iv) generating a refined labeled dataset that is based on the labeled dataset and a plurality of uncertainty scores associated with the plurality of validation classifications.

The invention relates to the technical field of machine learning potential molecular simulation, and provides a montmorillonite thermal conductivity prediction method based on NEP machine learning potential, and the method comprises the following steps: constructing a montmorillonite crystal structure model, carrying out first principle molecular dynamics calculation on the montmorillonite crystal structure model, introducing lattice distortion and atomic disturbance, and generating an initial structure set; performing static single-point DFT calculation on the initial structure set, and constructing a training data set; based on an NEP machine learning framework, utilizing the training data set to train and generate a potential function; optimizing the generation process of the potential function by adopting an active learning strategy; and performing thermal conductivity prediction verification on the potential function, and obtaining a thermal conductivity prediction result of the montmorillonite based on the potential function passing the verification. The defect that a traditional classical potential function depends on an empirical or semi-empirical model and is limited in precision and applicability is overcome, the problem that the dynamic behavior and the heat conduction mechanism of a complex material are difficult to accurately capture is solved, the calculation cost is reduced, and meanwhile the precision is kept.

The invention provides a water quality detection method based on combination of machine learning and spectrophotometry, and the method comprises the steps: measuring the absorbance of a water sample under a plurality of preset wavelengths through a spectrophotometer to form a multi-dimensional absorbance data set, mapping the multi-dimensional absorbance data set to a high-dimensional space through a kernel function technology to obtain a data set with enhanced non-linear features, and carrying out the detection of the non-linear features of the multi-dimensional absorbance data set. Extracting a known water sample and the pollutant concentration corresponding to the known water sample from historical water quality data, generating a training sample set, performing pollutant concentration prediction on the data set with the enhanced nonlinear characteristics by applying a machine learning model, and dynamically adjusting the training sample set by adopting an active learning strategy to obtain a pollutant concentration prediction value, and performing safety grade evaluation on the water sample and outputting a water quality report by using the predicted value of the pollutant concentration in combination with a water quality safety standard, the technical scheme provided by the invention improves the efficiency and accuracy of water quality detection, provides powerful technical support for environmental protection and public health guarantee, and has a wide application prospect. The whole process realizes integrated operation from data acquisition to result interpretation, and ensures scientificity and reliability of water quality management decision.

A computer implemented method of managing a first model that was trained using a first machine learning process and is deployed and used to label medical data. The method comprises determining (202) a performance measure for the first model, and if the performance measure is below a threshold performance level, triggering (204) an upgrade process wherein the upgrade process comprises performing further training on the first model to produce an updated first model, wherein the further training is performed using an active learning process wherein training data for the further training is selected from a pool of unlabeled data samples, according to the active learning process, and sent to a labeler to obtain ground truth labels for use in the further training.

The present invention relates to an AI-based system and method for generating enhanced radiology reports. The system comprises a database for storing multimodal patient data, a natural language processing (NLP) module for extracting clinical information, and a machine learning module for correlating the clinical information with radiology images to identify diagnostic insights. An AI-based report generation module analyzes the images and clinical information to generate a preliminary report, which is refined based on radiologist input. The generated report is then integrated into the patient's electronic health record. The system employs techniques such as multimodal deep learning, active learning, explainable AI, and federated learning to enhance diagnostic accuracy, capture expert feedback, provide transparency, and enable multi-institutional collaboration. The invention aims to improve the accuracy, efficiency, and value of radiology reporting in patient care.

Various embodiments of the present disclosure provide methods, apparatus, systems, computing devices, computing entities, and / or the like for labeling data by (i) generating, using a natural language machine learning model, a labeled dataset from unlabeled data, (ii) training one or more instances of a classification machine learning model based on the labeled dataset, (iii) generating, using the one or more instances of the classification machine learning model, a plurality of validation classifications, and (iv) generating a refined labeled dataset that is based on the labeled dataset and a plurality of uncertainty scores associated with the plurality of validation classifications.

Methods, apparatus, and computer-implemented methods are provided for training machine learning (ML) techniques to generate a property model for predicting whether a compound has a particular property. An iterative process / feedback loop can be performed to generate a property model, the process including generating, based on the property model, a list of prediction results for a plurality of compounds and their association with the particular property, validating the property model based on compounds from the list of prediction results having an association with the particular property, and updating the property model based on the property model validation. The process / loop can be repeated using the updated property model until it is determined that the property model has been effectively trained. The property model validation can include selecting a list of compound candidates, performing simulation analysis and / or laboratory analysis on the list of compound candidates for association with the particular property, and updating the property model using the simulation and / or laboratory results.

The application provides an alloy LIBS quantitative analysis method based on a Bayesian convolutional neural network and active learning. The method comprises the following steps: S1, spectrum data acquisition and preprocessing; S2, model construction and training; S3, active learning iteration optimization. Compared with a traditional machine learning model, the alloy LIBS quantitative analysis method provided by the application has high accuracy and good interpretability. Under the same number of labeled samples, the application shows better regression performance compared with a random sampling and a representative sampling strategy. Even if the number of labeled samples is reduced to 65% of the data set, the prediction performance of most components can reach 90% of full supervised learning, high accuracy is maintained, and good engineering application prospects are achieved.

This invention relates to a method for constructing a machine learning potential energy model based on hierarchical active learning, comprising the following steps: constructing a database of background solvent molecules and lithium salt molecules; initializing a machine learning potential energy model committee; outer active learning automatically exploring the chemical species composition space of the electrolyte, generating electrolyte formulations and initial structures; determining whether a structure needs to be included in the labeling range based on uncertainty calculations; automatically constructing a liquid phase environment based on classical force field simulation methods; performing high-precision simulations based on first-principles calculations and labeling the energy and force values ​​of the structures; inner active learning driving the machine learning potential energy model to explore the configuration space and label the energy and force values ​​of structures with high uncertainty; obtaining a first-principles structure database of the real liquid phase environment through several iterations; and training to obtain the final high-performance machine learning potential energy model. This invention systematically and comprehensively samples the chemical species composition space of the electrolyte and the molecular simulation configuration space based on hierarchical active learning, thereby ensuring the predictive performance and simulation accuracy of the machine learning potential model on unknown and complex electrolyte systems.

A method of datapoint elimination for an active learning algorithm includes monitoring datapoints in a labeled training dataset as new labeled datapoints are added to the labeled training dataset; determining whether datapoints in the labeled training dataset satisfy a criterion for elimination operations of an elimination protocol; and applying the elimination operations of the elimination protocol to remove one or more datapoints from the labeled training dataset.

A method of generating annotations corresponding to objects within images based on neural networks iteratively trained using the images. Active learning is used to train a machine learning model 256 b

The present invention discloses an active learning method and device, the method comprising: obtaining labeled sample data; training the labeled sample data using a machine learning model to obtain an intermediate model; inputting test data into the intermediate model for prediction to obtain a prediction result corresponding to the test data; outputting the test data corresponding to the prediction result with fuzzy boundaries as sample data to be labeled according to a preset evaluation index; wherein the machine learning model is trained and the intermediate model is used for prediction in a multi-threaded parallel manner. The present invention provides users with a highly versatile machine learning platform. The present invention can reduce the amount of labeling of sample data for input model training, reduce the labor cost of manual labeling, and effectively improve processing efficiency. The effect of the model can be quickly iterated and optimized at the minimum cost of labeled data. Minimized consumption of processing resources is achieved.

An exemplary embodiment of the present disclosure provides a system for generating a set of stimulation parameters, the system comprising at least one processor and a memory in communication with the at least one processor and having stored thereon instructions that, when executed by the at least one processor, is configured to cause the system to analyze, using a machine learning model, data from the deep brain stimulation device or an electromyography and generate, in response to analyzing, at least in part, the data, a set of stimulation parameters for deep brain stimulation.

Systems, methods, and devices for active learning are provided. The system includes a data storage device, a dataset analysis tool, a synthetic data generation module, an anomaly module, an image search engine module, an explainable AI module, an automated training platform, and optionally, a federated learning module. The system may be configured to operate on a general-purpose or dedicated computer and may further include a processor, memory, and a network interface. The system analyzes the provided dataset and generates synthetic data for augmenting the data within the provided dataset through the interaction of its components. The provided data and the generated data are used to train a machine learning model. The system may be operated iteratively and sequentially to continuously improve the machine learning model trained by the system by applying explainable artificial intelligence techniques with little or no human intervention.

The invention discloses an efficient deep batch active learning method and system based on gradient regularization, and relates to the technical field of machine learning and artificial intelligence, and the method comprises the steps: obtaining an unmarked sample, inputting the unmarked sample into a pre-established deep neural network model, and carrying out the output to obtain an output probability vector and a feature vector of a penultimate layer; according to the output probability vector and the feature vector of the penultimate layer, allocating a pseudo tag to the unmarked sample, calculating a gradient block of each category of the output layer based on the pseudo tag, performing regularization processing on the gradient blocks, retaining the gradient blocks corresponding to the category of the preset output probability, performing zero setting on the other gradient blocks, and generating a sparse gradient matrix; the method comprises the following steps: performing classification-level and sample-level hierarchical distance calculation on the basis of a sparse gradient matrix to obtain a sample-level distance, performing diversity sampling according to the sample-level distance, selecting representative samples by using a k-means + + clustering algorithm for labeling, and realizing acceleration and performance optimization of sample selection in deep batch active learning.

A computing system including one or more processing devices configured to compute a simulation of a chemical system. Computing the simulation includes, at a machine learning force fields (MLFF) model, computing local label sets of respective local regions. Each local label set includes MLFF labels. Computing the simulation further includes computing local uncertainty values and identifying local regions based on the uncertainty values. For each identified local region, computing the simulation further includes computing a substructure that surrounds that region. The substructure is selected by computing a substructure boundary that approximately minimizes a substructure uncertainty. Computing the simulation further includes selecting atomic positions within the substructure that approximately minimize the substructure uncertainty. Computing the simulation further includes computing substructure labels via quantum chemical simulation, training the MLFF model using the substructure labels, and outputting a potential energy surface (PES) specified by the MLFF labels and the substructure labels.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a machine learning model. In one aspect, a method comprises: determining a respective importance score for each mRNA codon in a set of possible mRNA codons; at each of a plurality of training iterations in a sequence of training iterations, selecting a current batch of mRNA sequences for training the machine learning model at the training iteration using the importance scores for the mRNA codons, and training the machine learning model on at least the current batch of mRNA sequences; and outputting the trained machine learning model.

The invention discloses a passive operating system identification method based on active learning. The method comprises the following steps: extracting fingerprints of network traffic by using Wireshark to obtain a fingerprint data set; the fingerprint data set is preprocessed, the operation system type, the IP signature and the TCP signature are separated, and the fingerprint data set is sampled and balanced through the combination of an oversampling algorithm BorderLineSMOTE and an undersampling algorithm TomeLink; the preprocessed fingerprint data set is input into a pre-trained fingerprint recognition model, a recognition result output by the fingerprint recognition model is obtained, and the fingerprint recognition model is obtained through active learning training; according to the passive operating system identification method based on active learning, the requirement for the number of data sets is lowered, the problem that the classification result of machine learning is biased to the operating system category with the large occupation ratio is solved, and then the identification accuracy is improved.

The invention provides an infertility patient pregnancy prediction method based on active learning and an integrated model, and the method comprises the following steps: carrying out the preprocessing of collected data, including classification variable coding, standardization processing and multi-dimensional abnormal value detection; generating a few categories of synthetic samples through an SMOTE technology, and optimizing data distribution in combination with a Tomek Links method; training a pregnancy prediction model based on machine learning models such as a random forest and XGBoost; the prediction results of the multiple basic models are fused through the Blending technology; based on an uncertainty sampling strategy, selecting samples for labeling and carrying out mixed training; and the pregnancy success prediction probability and credibility are output by inputting patient data. According to the method, the prediction accuracy and stability of the model can be improved, the robustness of the model to noise data and boundary samples is improved, the generalization ability of the model on new data is improved to a certain extent, doctors and patients can obtain more accurate treatment decision support, and the success rate and efficiency of IVF-ET treatment are improved.

A method of automated control of a microscope in cryogenic electron microscopy (cryo-EM), wherein the microscope is configured to collect high-magnification micrographs of particles suspended in vitreous ice. Such particles are found in grid squares, and a square contains holes from which high-magnification micrographs are imaged. The method is carried out during an active data collection session, leveraging a pipeline that comprises a set of models. The pipeline evaluates a set of collection locations to determine whether to continue collection at a current grid / square or instead at a new grid / square. The evaluation is based on a set of one or more quality scores derived from one or more pretrained models and machine learning-based active learning. Based on the determination, control information is provided to automatically control the microscope to move to a next target for data collection.