1508 results about "Classifier (UML)" patented technology

Disclosed in the present invention are an adaptive deep transfer fault diagnosis method and system, an apparatus and a medium. The method comprises the following steps: S1: collecting vibration acceleration signals of industrial equipment under different working conditions, and dividing same into a source domain data set and a target domain data set; S2: building a self-tuning universal domain adaptive fault diagnosis model, which comprises a shared feature extractor, a known classifier and a plurality of unknown classifiers; S3: separately calculating a classification loss of known faults of the source domain, a discriminative loss of the plurality of unknown classifiers, a target domain soft consistency regularization loss and an information entropy maximization loss; S4: introducing a dynamic weighting strategy based on model uncertainty assessment to optimize the model parameters; and S5: using the model for diagnosis. The present invention can fully mine valid information in data, can establish reliable class decision boundaries, and in addition, uses the self-tuning dynamic update strategy to adjust weightings corresponding to different loss functions, thus allowing for quick generalization of the model to different industrial diagnosis scenarios.

The invention provides a flow analysis and threat detection method and device based on machine learning, and the method comprises the steps: collecting a real-time flow data package of a target network environment, carrying out the protocol analysis and session recombination, and generating a real-time flow feature data set containing multi-dimensional flow features; loading a pre-trained multi-level threat classification model, inputting the real-time traffic feature data set into a feature extraction layer of the model, carrying out normalized coding on traffic features of corresponding dimensions through feature coding channels, generating a real-time feature vector sequence, inputting the real-time feature vector sequence into a primary classifier of the model, and classifying the real-time traffic features according to the real-time feature vector sequence; and performing abnormal probability calculation and cluster division on the real-time feature vector sequence through a mixed detection unit, outputting a primary threat tag and an abnormal confidence coefficient corresponding to each real-time feature vector, inputting the primary threat tag and the abnormal confidence coefficient into an aggregation classifier, performing dynamic weighted aggregation, and generating a comprehensive threat score so as to judge whether a threat response strategy is triggered or not. According to the invention, the accuracy and timeliness of threat detection in a complex network environment can be improved.

A database system generates a prompt for an LLM or other machine learning (ML) model to narrow the search space to highly relevant information about a database. A distinct instance of a classifier, a clustering algorithm, or a topic modeling model can be trained based on information from ML automation within the database system, respectively for each column or table in the database. Model instances can then be used during generative LLM inferencing to identify relevant sources of data to answer the user's question. Thus, the prompt generation combines ML automation and other ML models or an LLM for topic modeling and schema description.

The invention discloses a multi-modal fusion rumor detection method and system based on a dynamic graph convolutional neural network. According to the method, a dynamic feature graph of a language propagation path is constructed, and potential features in the language propagation process are extracted and analyzed by utilizing time sequence changes and key node relations between nodes in a propagation graph. A neural network is adopted to extract and enhance image data, text semantic features are extracted in combination with a text feature modeling network, text feature vectorization expression is achieved based on a BERT model, and rich semantic information is obtained. And a gating mechanism is introduced to dynamically adjust fusion weights of different modal features, and an information fusion strategy is optimized. A collaborative attention mechanism is further adopted for deep fusion, interactive learning of text, image and propagation path features is enhanced, and the relevance of cross-modal and time series data is improved. And finally, inputting the fused feature vectors into a classifier for accurate classification, thereby realizing accurate detection of the social media rumors. According to the method, the multi-modal features are effectively integrated, and the false information identification efficiency is remarkably improved.

Systems and methods for a prompt generation and analysis service for generating and identifying a preferred prompt for performing a function of a large language model (LLM) are provided. The prompt generation and analysis service may generate a set of training prompts for performing a function of an LLM. The prompt generation and analysis service may then query the LLM with the generated set of prompts and characterize the output of the LLM for each prompt. Using the characterization of the output and corresponding prompt, the prompt generation and analysis service can train a classifier model to classify the prompts. The prompt generation and analysis service may generate a set of target prompts for performing a function of an LLM, characterize the target prompts using the training classifier model, and identify a preferred prompt for performing the function based on the classifier model's classification.

The invention discloses a class incremental learning method based on elastic knowledge storage and compensation, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining images which comprise different tasks and are not intersected in class, constructing a data set of multi-task incremental learning, and dividing the data set into a training set and a test set; constructing a CIL model comprising a feature extractor and a classifier; wherein for the feature extractor, a pre-trained Transform model is given, and the model is made to adapt to a downstream class incremental learning task through a learning task sharing adapter; based on an importance perception parameter regularization framework and a semantic drift compensation framework, performing cross-task training on the CIL model by using the training set; and inputting the test set into the optimal CIL model to realize class incremental learning based on elastic knowledge storage and compensation. According to the method, effective forgetting resistance can be achieved without adding extra parameters, and meanwhile stability is kept and model plasticity is reserved in the training period.

The invention discloses a tensor depth semi-supervised learning method for high-dimensional small sample data classification. The tensor depth semi-supervised learning method comprises the steps of preprocessing original high-dimensional small sample data; constructing a deep neural network comprising a feature extraction module and a classifier module; constructing a second-order similarity matrix and a third-order similarity tensor based on the low-dimensional embedding representation obtained by pre-training; combining the second-order similarity matrix and the third-order similarity tensor to construct an objective function containing multi-order smooth constraints; then, taking the low-dimensional embedded representation obtained by pre-training as input, performing iterative optimization on the target function by adopting a gradient descent algorithm through a label propagation network formed by a full connection layer, and generating a pseudo label; and inputting the original high-dimensional small sample data, the low-dimensional embedded features and the pseudo labels into the deep neural network, iteratively updating the network in a semi-supervised mode until convergence, and outputting a final prediction result. According to the method, more accurate label propagation is realized, and the semi-supervised classification precision is improved.

Systems and methods for generating summaries from text using a generative model are disclosed. The system is configured to access an article; identify section; provide, to one or more generative models, a prompt including instructions to generate a section summary; generate an article summary based on the section summary; determine, from the article summary, a first concept found in the article summary that is missing from the article; determine, using a classifier, for a first sentence included in the article summary, a confidence score; and provide, for presentation at a client device, a document including the article summary.

The invention discloses an analysis method and system for a dynamic recrystallization structure form of a titanium alloy based on machine learning and a medium, and belongs to the technical field of metal material microstructure quantitative characterization, a Gleeble thermal compression test is processed on a titanium alloy to be tested, multi-modal data is collected, a DRX probability graph is output based on a DRX segmentation model, and then the characteristics of DRX are extracted. And respectively inputting the DRX features into the first-level classifier, carrying out PCA dimension reduction processing, splicing and fusing the prediction probability and the features after dimension reduction, and inputting the spliced and fused features into a second-level classifier to output a two-dimensional DRX segmentation map. Based on the FIB-SEM tomography sequence image, reconstructing a three-dimensional model of the DRX crystal grain so as to carry out consistency verification on the two-dimensional DRX segmentation image; and evaluating a correlation coefficient of the three-dimensional model and the two-dimensional DRX model, and finally carrying out three-dimensional visualization on the space aggregation of the DRX crystal grains. According to the method, the automation degree and efficiency of dynamic recrystallization proportion and type identification are remarkably improved.

The invention provides a self-supervised group behavior recognition method and system based on global and local comparative learning, and the method comprises the steps: extracting individual features through local branches, generating a soft mask by employing a multi-head self-attention mask module, separating significant / non-significant individual features through mask pooling, constructing a comparison sample, and carrying out the recognition of a group behavior through the comparison sample; feature alignment is optimized in combination with cosine similarity and local contrast loss; meanwhile, the spatial interaction relation of the behaviorists is captured by utilizing the spatial global Transform of global branches, short-term action and long-term behavior modes are fused through multi-scale time sequence coding, the consistency is optimized by adopting global comparison loss after spatial and temporal characteristics are aggregated, and finally, the two branch characteristics are integrated through global-local comparison loss for comparison, so that the accuracy of the behavior behavior is improved. And the loss weight is automatically adjusted to simplify parameter adjustment. After training is completed, a group behavior recognition result is output through the classifier according to the extracted spatial-temporal features, and the discrimination ability and robustness of the model to complex group behaviors are effectively improved.

The invention discloses a diffusion model image generation method and device for structural disturbance guidance and consistent distillation fusion, and the method comprises the steps: carrying out the initialization of a first student network through a diffusion model, and initializing a first teacher network; generating an initial state of the submerged space according to the submerged space data set and a noise scheduling function of the diffusion model; introducing the classifier-free guidance after the structural disturbance into a first teacher network to obtain a second teacher network; according to the first teacher network, the second teacher network and the initial state of the submerged space, state prediction is carried out through a solver, and enhanced input of the first teacher network is generated; obtaining consistency loss according to the first student network, the first teacher network and the enhanced input; according to the consistency loss, performing parameter updating on the first student network to obtain a target model; the target model is used for generating a target image. The method can improve the quality of the generated image and increase the generation speed of the diffusion model, and can be widely applied to the technical field of image generation.

The invention discloses an enterprise credit risk assessment method based on big data collection, and relates to the technical field of big data analysis, and the method comprises the steps: constructing a space-time fusion engine, carrying out the fusion through combining a cross-modal attention mechanism, generating a space-time fusion feature, and carrying out the adversarial training through a gradient inversion layer and a domain classifier, eliminating space-time fusion feature distribution differences; based on space-time fusion features, constructing a causal graph skeleton by adopting a conditional independent test algorithm, quantifying causal effect intensity among nodes through a machine learning model, constructing a risk conduction dynamic model, and predicting risk conduction intensity based on the causal effect intensity; and based on the risk conduction intensity, calculating a risk index weight through a dynamic game network, generating an enterprise risk score in combination with the space-time fusion feature, and generating a dynamic risk score through a time sequence neural network. According to the invention, by constructing the space-time fusion engine and combining the cross-modal attention mechanism, efficient fusion of multi-modal data is realized, and the accuracy and robustness of feature expression are improved.

The invention relates to the technical field of block chains, in particular to a smart contract vulnerability detection method and device based on multi-modal features, and the method mainly comprises the steps: training a meta-learning model in a dynamic adaptation module, and adjusting the global parameters of a modal feature extraction module, a dynamic gating fusion module and a classifier through the meta-learning model, the dynamic adaptation module comprises a meta-learning model constructed based on an MAML framework, and is used for optimizing global parameters of each module according to vulnerability features learned in pre-training; and inputting the multi-modal fusion feature vector into a classifier, and generating and outputting a vulnerability detection result of the smart contract. According to the method, known vulnerabilities can be accurately detected by fusing multi-modal features, and novel vulnerabilities can be rapidly adapted and detected.

Systems and methods for resource conservation based on query complexity are disclosed. An input query is received (e.g., via a chat interface) and provided to a response classifier, which is a machine-learning classifier that is trained to classify input queries with complexity scores that indicate how difficult it is likely to be for an artificial intelligence (AI) model to generate a response to the input query. If the complexity score exceeds a threshold score, the input query is provided to a first AI model (e.g., a relatively high-complexity AI model having a large number of parameters, relatively long response latencies, and / or other performance characteristics). If the complexity score does not exceed the threshold score, the input query is provided to a second AI model (e.g., a lower-complexity AI model having fewer parameters, shorter response latencies, and / or other differences in performance characteristics relative to the first AI model).

The invention discloses a wind turbine generator equipment fault diagnosis method and system based on a large model. The method comprises the following steps: initializing a feature set according to normal data; selecting a plurality of candidate features according to the mixed score of each candidate feature, and storing the candidate features in a feature set; performing sample division on normal data by using a time sequence segmentation algorithm and constructing multi-dimensional spatial-temporal feature representation; constructing a fault diagnosis model: loading a large language model as an infrastructure, and injecting the multi-dimensional spatio-temporal feature representation into an embedded input layer of the large language model; an adaptive pooling layer is accessed after the output of the large language model, and a double-layer MLP classifier is constructed for realizing the identification and classification of different fault types; the first layer of the double-layer MLP classifier compresses an input feature to half of an original feature dimension and applies GELU activation, and the second layer of the double-layer MLP classifier is mapped to a corresponding fault category space; constructing a knowledge mechanism library, providing prior knowledge of the wind turbine generator for the model, designing a loss function driven by the knowledge of the wind turbine generator, and carrying out model training.

The invention discloses a large reasoning model factuality enhancement method, system and device based on reinforcement learning and a medium. The method comprises the following steps: constructing a large language model reasoning thinking chain factuality detection and enhancement framework (REELANCE), and performing fine-grained fact accuracy evaluation on a reasoning chain output by a large language model; based on the result of the fact check classifier, the reasoning chain factuality is enhanced by adopting a reinforcement learning group strategy optimization technology (GRPO); a multi-dimensional reward mechanism is designed to improve the fact accuracy of the reasoning step; quantitatively analyzing the influence of enhanced training on the nerve activation change in the model through a mechanism interpretability technology; a comprehensive optimization target giving consideration to the fact accuracy and the reasoning quality is established, the fact robustness is improved, the standard reasoning performance is kept, and the balance between the fact and the reasoning ability is achieved; the key problem of inaccurate facts of an inference type large language model in a high-risk scene can be effectively solved; the invention also provides a system, equipment and a medium for realizing the method.

The invention relates to the technical field of unmanned aerial vehicle fault diagnosis, in particular to an unmanned aerial vehicle cross-working-condition fault diagnosis method based on double-alignment federated metric learning. Comprising the steps that a server initializes a global model; the client calculates classification loss based on local data, optimizes measurement learning loss and feature alignment loss by using clustering anchor points and unbiased anchor points, updates a local model, and optimizes a classifier by calculating the alignment loss of a classifier through the unbiased anchor points; the client side calculates a feature anchor point and uploads the anchor point and a local model to the server; the server clusters the feature anchor points by using a parameter-free clustering algorithm to obtain clustering anchor points of each class, averaging the anchor points in each clustering cluster to obtain unbiased anchor points of each class, performing weighted averaging on all local models, and updating a global model; and repeating the above process iteration until convergence. According to the invention, through a double-alignment mechanism and federated metric learning, the problem of feature isomerism and the challenge of poor model generalization ability in unmanned aerial vehicle fault diagnosis are solved.

Hierarchical modelling and advanced feature engineering discover abnormalities in time series data with irregular trends. Data is collected in real time to ensure temporal integrity in the invention. Extraction filters and isolates useful data. Data cleansing removes noise and extraneous data after preliminary analysis identifies patterns and abnormalities. Feature engineering organizes cleansed data for machine learning algorithms. Primary storage stores this data for fast retrieval and extensive trend analysis. Holidays and weekends provide unique patterns in trend analysis. These trends are used to cluster data and create hierarchical predictive models, starting with a first-order model for general trends and increasing in order to refine residuals. Serializing these models improves storage and retrieval. Trend clusters are created from new data points, and algorithms detect pattern deviations. Statistical tests and machine learning classifiers identify anomalies and create alerts and remedial measures. The system monitors and analyzes incoming data to detect anomalies.

The invention provides an unknown encrypted traffic identification method and system based on small sample incremental learning, and a storage medium, and the method comprises the steps: 1, carrying out the fine-grained classification of known traffic: extracting the features of each level of original encrypted traffic, building a respective variational auto-encoder for each type of known traffic, generating a potential representation, and carrying out the fine-grained classification of the known traffic; inputting into a classifier to classify known attacks in a fine-grained manner; step 2, specific label distribution of unknown traffic: judging whether the sample is a drift sample or an unknown sample by adopting a scoring function, and performing hierarchical clustering on the samples according to each hierarchical feature to realize label distribution of the unknown traffic; and step 3, dynamically updating the classification model: training a new classifier by adopting a new sample, connecting other classifiers to form a classification graph, and updating nodes of the classifiers by adopting a graph attention network to realize small sample incremental learning. The method has the beneficial effects that low-sample incremental modeling of a new class is effectively supported, and the fine-grained recognition capability and the model generalization adaptability are remarkably improved.

An analysis engine receives data characterizing a multimodal prompt for ingestion by a generative artificial intelligence (GenAI) model. The multimodal prompt is processed and fed into a plurality of layers from which an intermediate result of the GenAI model or a proxy of the GenAI model is obtained. The analysis engine, using a prompt injection classifier and the intermediate result, determines whether the prompt comprises or is indicative of malicious content or elicits malicious actions. Data characterizing the determination is provided to a consuming application or process. Related apparatus, systems, techniques and articles are also described.

The invention discloses an autoregressive language model training process-oriented data leakage risk quantification method, which comprises the following steps of: executing enhanced member data division processing on an original training text set, and generating a ternary partition text set containing member, non-member and key boundary texts through an optimization function; for each text, using the target model to extract member attributes from three channels of forward reasoning, back propagation and state evolution, and fusing the member attributes into member attribute vectors; inputting the member attribute vector into a hierarchical comparison embedded network, and mapping the member attribute vector into an optimized embedded representation vector through comparison learning including similar clustering, heterogeneous separation and boundary positioning; and inputting the embedded representation vector into a dynamic reasoning classifier capable of perceiving a training stage, judging the risk membership degree of the dynamic reasoning classifier, and generating a data leakage risk quantification result. According to the invention, real-time and fine-grained dynamic quantification can be carried out on the leakage risk, and timely early warning is provided for model training.

The invention relates to the technical field of fault prediction and computer big data processing, in particular to a generalized zero sample composite fault diagnosis method, device and system based on anti-fact reasoning. According to the generalized zero sample composite fault diagnosis method based on the anti-fact reasoning, a two-stage generalized zero sample composite fault diagnosis model based on the anti-fact reasoning is constructed. According to the model, internal causal components of fault data are pointed out from the angle of causal theory, and then a structural causal model is constructed to describe decoupling and generation of fault features under the guidance of anti-factual reasoning. On the basis, a generative model is improved through a reinforced discriminator in the first stage so as to realize binary classification of a single fault and a composite fault. In the second stage, a single fault category is predicted through supervised training of a classifier, and meanwhile, a traditional zero sample learning method is designed to classify composite faults. According to the method, the diagnosis precision of the model is greatly improved, and the problem of deviation of model diagnosis on visible classes and invisible classes is solved.

The invention discloses an incomplete multi-view multi-label data classification method based on semantic enhancement and pseudo-label uncertainty perception, and the method comprises the steps: employing a dual-channel feature extraction and decoupling module to obtain the shared semantic representation and specific representation of each view in each sample for a constructed incomplete multi-view multi-label data classification network model; performing cross-view fusion on the shared semantic characterization and the specific characterization, obtaining a unified shared characterization and a unified specific characterization corresponding to each sample, performing feature fusion, obtaining a fusion characterization of each sample, inputting the fusion characterization of the sample output by the dual-channel feature extraction and decoupling module into a classifier for multi-label prediction, and performing multi-label prediction on the fusion characterization of the sample. Therefore, a multi-label classification prediction result is obtained, and model training is carried out based on a total contrast learning loss function and a joint supervision classification loss function. According to the method, the classification performance and the model robustness on incomplete multi-view multi-label data are remarkably improved through training learning under the guidance of semantic enhancement and uncertainty.

Techniques are described herein for a method of obtaining a token based on a conversation in real time. The method further includes predicting, using a large language model (LLM) and the token, a next token. The method further includes predicting, using a classifier and the next token, a completion of a user turn. The method further includes triggering a next turn of the conversation in real time using the completion of the user turn.

The invention relates to a food reserved sample quality monitoring method based on machine learning, and the method specifically comprises the following steps: deploying a sensor array in a food reserved sample environment, collecting reserved sample multi-dimensional physicochemical characteristic data in real time, constructing a training set through historical normal and degraded batch data, and marking a state; the method comprises the following steps: constructing a machine learning model for food reserved sample quality monitoring, inputting sample data in a training set into the model, sequentially passing through a dynamic distribution alignment module, a physical constraint confrontation enhancement module and a multi-scale residual space-time network, and performing dynamic attention mechanism enhancement by utilizing physical constraint and environment modulation. And finally, outputting a quality category probability by the multi-modal classifier. Then calculating model loss, and carrying out iterative training on the model to obtain a trained model; and deploying the trained model at a reserved sample monitoring terminal, inputting newly collected monitoring data, and predicting the quality state of the food reserved sample. According to the invention, high-efficiency and real-time quality monitoring on the quality of the reserved food sample can be realized.

The invention discloses a water conservancy project tunnel construction safety assessment method. The method comprises the steps that monitoring data of a tunnel are collected, and a safety assessment level is marked; expansion of monitoring data is carried out through a historical iteration generative adversarial network; inputting the monitoring data into a feature extraction model to train the feature extraction model, wherein the feature extraction model is a full-connection neural network based on dynamic adaptive oscillation; inputting the monitoring data into a feature dimension reduction model to train the feature dimension reduction model, wherein the feature dimension reduction model is a self-encoding neural network based on feature refinement; inputting the monitoring data into a classifier for training a classifier model, wherein the classifier is a fractional order neural network based on dynamic pruning of a fractional rank; and processing and classifying in the trained feature extraction model, feature dimension reduction model and classifier model to obtain a classification result. According to the invention, the problem of poor generalization ability caused by insufficient data samples in the existing engineering construction safety assessment method is solved.

The invention discloses a karst stratum tunnel disaster sensing method and system based on multi-source information, and belongs to the technical field of tunnel engineering safety monitoring, and the method comprises the steps: building a monitoring index data set, converging the monitoring index data set to a cloud end, carrying out the time-space registration, and generating a multi-dimensional time sequence data field; calculating data uncertainty of each monitoring area by adopting an information entropy theory, calculating spatio-temporal evolution characteristics, and performing classifier identification by combining a deformation field space gradient to obtain a key monitoring area; adaptively adjusting the acquisition frequency of the sensor, and starting supplementary monitoring equipment for encrypted observation; performing space-time response calculation by adopting a machine learning algorithm to generate a tunnel disaster evolution prediction result; and carrying out grading threshold comparison and numerical simulation verification on the prediction result to realize effective identification and perception of the disaster evolution state. According to the method, the technical means of combining multi-source data fusion, the information entropy theory, machine learning and self-adaptive monitoring is adopted, and dynamic, accurate and predictive perception of the tunnel disaster evolution process can be achieved.

Systems and techniques for are described herein. A content passage is received from a corpus of training data comprising labeled training data and unlabeled training data. A query is received from a query hierarchy for a classification domain. The content passage and the query are embedded to form a passage-query pair. A predicted result for the passage-query pair is generated based on a calculated probability of the predicted result being within an answer threshold. A passage-query-result triplet is generated that comprises the passage-query pair and the predicted result according to the query hierarchy for the classification domain. Vectors of the content classification large language model are updated using the passage-query-result triplet.

The present invention relates to a computer-implemented multi-module classifier method and system for providing a pathogenicity classification score of a variant of a protein of interest. The classifier comprises a sequence module based on a protein language model (PLM); a structure module based on a graph neural network (GNN); a property module; and a unified head module based on a machine learning model. The invention further relates to methods for preparing, training, and implementing the multi-module classifier system.

Typical classifiers must be trained on a large input sample to accurately classify inputs. In addition, if a new classification category needs to be added to a taxonomy after the classifier has already been trained to classify within the taxonomy, the classifier must be recreated and retrained to classify within the updated taxonomy. To address at least these technical problems with classifiers, a generative language model may be used to perform classification. A generative language model is a machine learning model that generates language, typically in the form of a textual response to a data input. A generative language model may utilize a large neural network to determine probabilities for a next token of a sequence of text conditional on previous or historical tokens in the sequence of text. An LLM is an example of a generative language model.