112 results about "Score" patented technology

Systems and methods for forecasting data drift for model monitoring. In some aspects, the system receives a current explainability vector for a machine learning model and a data drift vector for historical data profiles. The machine learning model is trained on historical data including values for a first set of features. The system generates a projected synthetic dataset using the data drift vector and updates the machine learning model based on the projected synthetic dataset. Using the current explainability vector and a future explainability vector for the updated model, the system generates a second set of features and determines a drift threshold vector for the second set of features based on values in the explainability vectors. The system determines a discrepancy score for each feature of the second set of features. The system generates an alert including features in the second set of features and their associated discrepancy scores.

The invention provides an unsupervised industrial anomaly detection method based on an improved full-convolution cross-scale flow network. The method comprises the steps of performing data preprocessing on an industrial image data set; industrial image data is used as input, a pre-trained visual backbone network is used for extracting multi-scale features, a resolution-perceived channel attention module is used for enhancing expression of the multi-scale features, and an enhanced multi-scale feature map is obtained; performing multilayer reversible transformation by taking the enhanced multi-scale feature map as input and taking an ICSF-Net model based on hierarchical attention and expansion convolution as a cross-scale normalized flow network, and modeling multi-scale feature distribution; in the training stage, optimizing and updating ICSF-Net model parameters by maximizing the log likelihood of a normal sample under potential Gaussian distribution based on multi-scale feature distribution and combining an LSGR mechanism; in the reasoning stage, probability density estimation is carried out based on multi-scale feature distribution, an abnormal score graph is generated, and defect detection and positioning are achieved. According to the invention, the accuracy, robustness and pixel-level positioning precision of industrial product defect detection are improved.

The invention discloses an SVG valve hall cooling efficiency evaluation method and system based on a probabilistic graph model, and the method comprises the steps: obtaining original time sequence data which is obtained through the collection of a cooling system multi-parameter monitoring sensor group disposed in an SVG valve hall in continuous T sampling periods; preprocessing the original time series data to obtain a credible time series data set; constructing a Bayesian network topological structure comprising three-level nodes of an environment layer, a component layer and an efficiency layer and causal dependence edges, and optimizing parameters of the Bayesian network topological structure by adopting a maximum likelihood estimation method to form a dynamic Bayesian network model after parameter calibration; and the credible time sequence data set is used as an evidence variable to be input into the Bayesian network model after parameter calibration, calculation is carried out through a belief propagation reasoning algorithm, a final control instruction set is generated through probability weighted scoring processing, the final control instruction set is fed back to a valve group monitoring system, and early warning and automatic load reduction are achieved. The problems of large evaluation deviation and early warning lag in the prior art are solved.

The invention discloses a large model fine tuning method and system based on a suffix prompt learning method and ensemble learning, and the method comprises the steps: carrying out the preprocessing of an existing data set, obtaining K subsets through the replacement sampling, enabling each subset to account for a specified percentage of the total, and constructing an artificial prompt; according to the teacher model and the student model, a predicted target word classification score is obtained through suffix prompt learning, through probability normalization, teacher knowledge is distilled to the student model by using a KL divergence loss function, and a prompt vector is initialized; an input task vector is obtained through context learning, and knowledge of the input task vector is distilled to a student model prediction vector by using a mean square error loss function; constructing a mixed loss function, and optimizing knowledge distillation and task feature learning; performing end-to-end optimization on the initialized prompt vector by using suffix prompt in combination with a negative log-likelihood loss function; during reasoning, after the optimized prompt vectors are spliced to the input vectors, suffix prompts share the input vector key value cache to generate output. According to the method, the adaptability and reasoning performance of the model are remarkably improved.

The invention provides a gushing prediction method and system for grouting engineering, and belongs to the technical field of tunnel and underground engineering construction safety monitoring. The method comprises the following steps: collecting grouting pressure, flow, slurry viscosity and micro-seismic event data; calculating a slurry resistance index and a micro-seismic event aggregation density; inputting the initial gushing risk index into a pre-trained machine learning model to obtain an initial gushing risk index and a confidence score; according to whether the confidence score is lower than a preset threshold, selecting direct early warning or entering a parameter correction path; when the confidence coefficient is low, dynamic correction factors are generated based on scores and risk changes, the calculation weight of the slurry resistance index is adjusted, corrected parameters are obtained, and secondary prediction is carried out; and integrating the initial prediction result and the secondary prediction result to obtain a final risk index, and executing graded early warning. According to the invention, the accuracy and reliability of gushing risk prediction are improved through a confidence-driven double-path decision and parameter on-line adaptive correction, and a constructed early warning mechanism.

The invention discloses an error detection method and system in a programmed task video, relates to a video time sequence action segmentation and anomaly detection technology, and realizes robust error detection aiming at the problem of accuracy of error recognition in the programmed task video. According to the method, firstly, a time sequence action segmentation model is trained, a Gaussian mixture model is trained for each action category by utilizing each frame feature of an intermediate layer of the time sequence action segmentation model, the frame features are sent to the Gaussian mixture model of the corresponding action category to be processed, the log-likelihood score of the frame is obtained, and the detection of the frame level anomaly is realized. According to the method, the problem of error detection in the first-view programmed task video can be effectively solved by combining action segmentation and anomaly detection methods, and the method can be used for scenes requiring accurate recognition of sequence action errors.

The invention provides a key point detection method based on deep learning, and the method comprises the steps: designing a lightweight network structure which comprises a feature encoder, a multistage feature fusion module, a reliability detector, a repeatability detector and a descriptor extractor, and outputting a reliability score graph, a repeatability score graph and dense descriptors; a non-maximum suppression sampling strategy is designed and fused, coordinates of a limited number of key points are sampled, bidirectional negative logarithm likelihood loss and bidirectional L1 distance loss are proposed to carry out coupling training on descriptors and reliability of the key points, and local similarity loss and local peak loss are proposed to optimize the positions of the key points. Compared with the prior art, the method not only achieves the performance equivalent to that of the most advanced SOTA method in feature matching, but also greatly shortens the reasoning time.

According to embodiments of the disclosure, methods, apparatuses, devices and medium for model performance evaluation are provided. A method includes: applying, at a client node, a plurality of data samples to a prediction model respectively to obtain a plurality of predicted scores output by the prediction model, the plurality of predicted scores indicating respectively predicted probabilities that the plurality of data samples belong to a first category or a second category; determining values of a plurality of metric parameters related to a predetermined performance indicator of the prediction model based on the plurality of predicted scores and a plurality of ground-truth labels of the plurality of data samples; performing perturbation on the values of the plurality of metric parameters to obtain perturbed values of the plurality of metric parameters; and sending the perturbed values of the plurality of metric parameters to a server node.

According to various embodiments, described herein are systems and methods for training machine learning (ML) models to generate real-time scores to predict the probabilities of task completion. In one embodiment, an exemplary method includes the operations of receiving, from a data store, a set of features and a workflow for training a first type of ML models, the workflow specifying a data source, a number of stages and associated parameters for training the ML models; retrieving, from the data source, training data for the set of features; and segmenting the training data into different segments. The method further includes the operations of training a separate first type of ML model using each of the different segment of the training data in accordance with the workflow; and persisting the first type of trained ML models into the data storage. The method also includes using a trained ML model to generate probability scores and displaying the scores to users in real-time.

The invention discloses an enterprise ESG score prediction method based on machine learning, and the method comprises the steps: collecting multi-dimensional ESG, financial and risk data, carrying out the normalization and missing / abnormal value processing, and dividing a data set; a multi-model system is constructed, and the optimal LSTM is selected by using five-dimensional indexes to predict ESG rating; in combination with an LSTM result, by taking maximization of an ESG mean value-minimization of a fluctuation ratio as a target, GRU is used for early warning of a short-term risk, LSTM controls a long-term trend, after constraint is added, an optimal investment portfolio weight is output through PSO, and a model updating and portfolio rebalance mechanism is set. According to the method, the ESG prediction precision is improved, the ESG-investment optimization integration is realized, and the data timeliness demand is adapted.

The invention relates to the field of data processing, in particular to a carton production line monitoring method and system, and the method comprises the steps: collecting and preprocessing multi-dimensional time sequence data of a production process; constructing a standard hidden Markov model (HMM) based on historical normal data, and defining the hidden state of the HMM as a microscopic operation mode under a macroscopic process; extracting procedure context features, and constructing a procedure conformity evaluation model (PCAM) to calculate a conformity score; dynamically adjusting the emission probability of the standard HMM in combination with the score to form an improved IHMM; and calculating the log-likelihood probability of the real-time observation sequence by using the improved IHMM, and comparing the log-likelihood probability with a preset threshold to judge the production abnormality. According to the invention, the accuracy and reliability of abnormity monitoring can be effectively improved.

The invention relates to the technical field of data processing, in particular to a threat recognition system and method.The method comprises the steps that a self-adaptive collection module obtains network flow data and operation log output feature vectors; the multi-dimensional rule module generates rule confidence; the statistical modeling module outputs an abnormal probability score; the machine learning module outputs a classification probability in response to the abnormal probability score dynamic selection model combination; the simulation decision module is fused with the multi-source data to execute threat diffusion simulation and output hazard indexes; the grading execution module triggers grading response according to the hazard index; and the feedback optimization module adjusts model parameters in a cross-layer manner. According to the method, noise interference is eliminated through feature optimization, a multi-dimensional model is fused to calibrate an output result, a threat grading mechanism preferentially responds to key threats, and closed-loop feedback is continuously injected into a false alarm data training model, so that the problem of high false alarm rate caused by insufficient training data coverage of a machine learning model in a complex dynamic network environment is solved; security resource allocation is optimized and key threat response delay is shortened.

The invention belongs to the technical field of data processing, and particularly discloses a building outer wall hollowing evaluation method and system, and the method comprises the steps: obtaining a target building outer wall detection signal, carrying out the time window segmentation and normalization preprocessing, and inputting a self-encoder model pre-trained by a normal wall detection signal to obtain a reconstruction signal; calculating a residual error of the detection signal and the reconstruction signal, and extracting a multi-dimensional feature formed by residual error energy, a residual error maximum absolute value, peak density and a multi-scale feature score from the residual error; and substituting the multi-dimensional features into an abnormal scoring function to obtain an abnormal scoring value, comparing the abnormal scoring value with a threshold value based on normal sample statistical features, and judging whether hollowing or potential hollowing exists in the target outer wall area. According to the invention, tiny hollowing signals can be accurately captured, the detection objectivity and accuracy are improved, and the refined evaluation requirements of building outer wall hollowing are met.

The invention provides an adversarial attack sample generation method and related equipment, which are applied to the field of machine learning, and are used for determining an initial value of adversarial disturbance distribution of a diffusion model, determining a disturbance score of the adversarial disturbance distribution based on the initial value and an established classifier misleading evaluation function, and generating an adversarial attack sample. And updating the model parameters of the diffusion model based on the disturbance score until the disturbance score of the adversarial disturbance distribution corresponding to the diffusion model is not less than a preset score threshold, determining the current diffusion model, generating an adversarial disturbance sample according to the model parameters of the current diffusion model, and further generating an adversarial attack sample. According to the method for generating the adversarial attack sample, a plurality of data samples do not need to be acquired, the adversarial disturbance distribution of the diffusion model is adjusted, the disturbance sample is generated when the adversarial disturbance distribution meets the set condition, and the adversarial attack sample is generated based on the disturbance sample, so that the sample acquisition is fast, and the attack efficiency of implicit adversarial attack is improved.

The invention provides a watermark detection method and device based on Bayesian detection. The method comprises the steps of obtaining a target text to be subjected to large model watermark detection and a cue word corresponding to the target text; processing a token sequence spliced by the cue word and the target text through a language model to obtain probability distribution output by the language model for each token position; for the jth token in the T tokens of the target text, selecting the first k tokens before the jth token to be input into the hash function, and obtaining a random seed of the jth token; dividing a word list of the large model into a preferential selection set and a non-preferential selection set based on random seeds; on the basis of preset watermark bias, preferentially selecting the set and the probability distribution, and generating probability distribution after disturbance of the jth token; calculating a log-likelihood ratio by using the probability distribution before and after disturbance, and accumulating the log-likelihood ratio to a detection score of the current text; and if the detection score is higher than the threshold value, judging that the target text has the watermark of the large model.

A computer-implemented method for generating a classifier, comprising: receiving aggregated statistics objects, wherein the aggregated statistics objects comprise bin frequencies F0 and F1, wherein the bin frequencies F0 and F1 are calculated for each of a plurality of predictive features, conditioned on a target value being 0 or 1, respectively; bin-level covariances C0 and C1, wherein the bin-level covariances C0 and C1 are calculated for each pair of bins, conditioned on the target value being 0 or 1, respectively; feeding the aggregate statistics objects F0, F1, C0, and C1 into the classifier, wherein the classifier generates a score calculated as a sum of a plurality of flexible nonlinear shape functions applied to the plurality of predictive features, respectively; and training the classifier by fitting the plurality of shape functions to maximize a divergence for score separation between target value 0 and target value 1.

The present disclosure describes a patent management system and method for remediating insufficiency of input data for a machine learning system. A prediction to be performed is received from a user input. Relevant input data is determined to perform the prediction. The relevant input data is determined by applying filters based on the prediction to be performed. Prediction is performed by generating a plurality of predicted vectors. A confidence score for the generated plurality of predicted vectors is determined. If the confidence score is less than a predetermined threshold, the prediction is unreliable. The input data is expanded by gathering additional input data. The input data is expanded with the additional input data until the confidence score exceeds the predetermined threshold. A predicted output is generated with the expanded input data. The prediction output and the confidence score are provided for rendering.

Systems, methods, non-transitory computer-readable media to perform operations associated with the storage medium. One system includes a storage medium and an encoding / decoding (ED) system to perform operations associated with the storage medium, the ED system being configured to process a set of log-likelihood ratios (LLRs) and a syndrome vector to obtain a set of confidence values for each bit of a codeword, estimate an error vector based on selecting one or more bit locations with confidence values from the set of confidence values above threshold value and applying hard decision decoding to the selected one or more bit locations, calculate a sum LLR score for the estimated error vector, and output a decoded codeword based on the estimated error vector and the sum LLR score.

This application provides a model training method, training system, and related equipment. The method includes the following steps: acquiring positive sample data, wherein the positive sample data includes input samples and positive labels, the input samples include questions, and the positive labels include the answers corresponding to the questions; inputting the input samples into a first large model to be fine-tuned, obtaining the first negative label corresponding to the input sample; determining the first score of the first negative label based on the relevance of the first negative label to the question and the quality of the response to the first negative label based on a scoring model; inputting the input samples into the first large model; and fine-tuning the first large model based on a loss function to obtain a second large model. The loss function is used to guide the first large model to improve the prediction probability of positive labels and the prediction probability of negative labels with scores greater than a threshold, so that the model can not only learn standard answers that meet user expectations, but also generate more diverse answers, thereby improving the fine-tuning effect of the large model.

The invention relates to a named entity recognition method based on label embedding and feature extraction, belonging to the technical field of named entity recognition methods; the technical problem to be solved is: to provide an improvement of the named entity recognition method based on label embedding and feature extraction; the technical scheme adopted to solve the above technical problem is: firstly, a text sequence is defined, and each word is encoded by a pre-training layer of BERT to obtain an embedded representation of each word; then, a word sequence context representation is performed on the recognized text, a Bi-LSTM model is adopted to extract local context features, a label vector is adopted to classify context feature vectors in the document when performing document-level representation, and when decoding and predicting the recognized text, a sequence feature generated by a second-layer Bi-LSTM model is linearly mapped to an emission score of each label, the emission score is combined with a state transition score learned by a conditional random field, the probability of the label sequence is calculated, and finally a log-likelihood function is adopted to calculate the loss: the invention is applied to named entity recognition.

The invention discloses a load event detection method based on mixing of a rank likelihood ratio and physical characteristics, and relates to the technical field of intelligent power grids and non-intrusive load monitoring. The method comprises the following steps: acquiring an original aggregated power signal, and performing multi-scale preprocessing on an active power sequence to extract core power and trend components; a dual-path mixed discriminant variable is constructed, a statistical path calculates a rank likelihood ratio statistic by using non-parametric rank transformation, and a physical path calculates a physical difference score through logarithm space mapping; sensing an average power level in a current window in real time, dynamically adjusting a fusion weight factor of a statistical path and a physical path, and generating a mixed sensing score; performing state transition monitoring on the mixed perception score by using a finite-state machine with a time sequence memory feature, and locking preliminary candidate event points; and finally, gradient fine trimming positioning and physical median consistency verification are performed on the candidate points, pseudo events are eliminated, and a legal load event sequence is output. According to the method, the dependence on Gaussian distribution hypothesis is eliminated from the mathematics essence, the robustness to non-Gaussian interference is enhanced, the capture sensitivity for weak load and overlapping events is improved, the calculation overhead is low, and the real-time monitoring requirement of the edge side is met.

Systems and methods for optimizing multi-camera multi-entity artificial intelligence tracking systems. Visual and location information of entities from video feeds received from multiple cameras can be obtained by employing an entity detection model and re-identification model. Likelihood scores that entity detections belong to an entity track can be predicted from the visual and location information. The entity detections predicted into entity tracks can be processed by employing combinatorial optimization of the likelihood scores by identifying assumptions from the likelihood scores, entity detections, and the entity tracks, filtering the assumptions with unsatisfiable problems to obtain a filtered assumptions set, and optimizing an answer set by utilizing the filtered assumptions set and the likelihood scores to maximize an overall score and obtain optimized entity tracks. Multiple entities can be monitored by utilizing the optimized entity tracks.

The invention relates to threshold estimation and calibration for anomaly detection. This document is machine learning (ML) and extreme value theory (EVT) techniques for normalizing and thresholding anomaly scores without assuming value distributions. In an embodiment, a computer receives a number of non-normalized anomaly scores and selects a highest subset of the non-normalized anomaly scores that exceeds a tail threshold according to a peak over-threshold (POT). Parameters of a probability density function are trained according to EVT based on the highest subset of non-normalized abnormal scores. After training and in a production environment, a normalized anomaly score is generated based on the non-normalized anomaly score and trained parameters of the probability density function. Anomaly detection compares the normalized anomaly score to an optimized anomaly threshold.

The invention discloses a large language model multi-dimensional preference alignment method based on multi-attribute collaborative optimization, and the method comprises the steps: generating a multi-candidate response based on reference strategy sampling, constructing a preference data set containing main and auxiliary attributes through an attribute scoring mechanism, and precisely capturing human implicit multi-dimensional preferences; a learnable prompt pool and an expert group are constructed, after expert groups are divided through attribute combination, multi-dimensional features of samples are fused through a routing gating network, accurate matching of the samples and the experts is achieved, and different attribute combination modeling requirements are met; constructing a self-normalization importance sampling weight based on the log-likelihood difference of the current model and the reference model, and estimating and maximizing the joint expectation score of the primary and secondary attributes; and meanwhile, flexible control is carried out on the expected constraint violation condition of the secondary attribute, so that collaborative optimization of the primary and secondary attributes is realized. According to the method, the large language model better fits human multi-dimensional real preferences, performance does not need to be sacrificed in any attribute dimension, and the model output quality and scene adaptation capability are remarkably improved.

This invention relates to the field of fluid mechanics and transient process analysis of hydraulic systems, and discloses a method for determining the occurrence boundary and risk level of liquid column separation. The method utilizes a visual test bench to conduct multiple experiments under different combinations of test parameters, collecting data and simultaneously recording the liquid column separation state. Based on Buckingham's π theorem, dimensional analysis is used to construct multiple independent dimensionless parameter sets from the test parameter combinations. These are used as input features, with the separation state as the target variable. A machine learning algorithm is introduced for multiple batch training to calculate importance scores, selecting the dominant dimensionless parameter with the highest and most stable score. Subsequently, regression analysis is used to fit the data boundary, solving for the intersection point and outputting the critical equilibrium point as a quantitative benchmark for the occurrence boundary. After confirming separation, high-scoring dimensionless parameters are selected to construct composite judgment parameters, which are compared with preset thresholds to classify the risk level. This invention can accurately define the critical equilibrium point of cavitation occurrence and quantitatively assess its risk severity.

The invention discloses a voiceprint comparison method and device based on log-likelihood ratio calibration and a readable medium. The voiceprint comparison method comprises the steps of collecting voice data of multiple persons in a target scene and forming a voice data set; performing voiceprint comparison on every two pieces of voice data of the same person or different persons in the voice data set, and constructing homologous data and non-homologous data; respectively calculating homologous distribution and non-homologous distribution based on the homologous data and the non-homologous data, and obtaining a sampling score set and a log-likelihood ratio set; calculating a log-likelihood ratio calibration expression based on the sampling score set and the log-likelihood ratio set; and obtaining to-be-detected voice data and sample voice data in the target scene, performing voiceprint comparison to obtain a corresponding comparison score, and performing calculation according to the log-likelihood ratio calibration expression to obtain a corresponding log-likelihood ratio so as to determine the similarity between the to-be-detected voice data and the sample voice data. According to the invention, the application range and accuracy of the original calibration algorithm are improved.

This invention relates to the field of financial risk assessment technology, specifically to a case risk assessment system and method based on machine learning algorithms. The system mainly includes a pre-training module and a prediction module. This invention is designed with the application of artificial intelligence algorithms in the risk assessment of suspicious cases as its core concept. Guided by theories such as supervised learning and unsupervised learning, it analyzes the performance of various features of cases within a fine-grained feature system through classification algorithms, anomaly detection algorithms, and statistical analysis methods. This process uncovers hidden and anomalous information, assesses the risk of cases, and assigns scores and rankings them. Through the application of AI technology, financial institutions can more quickly and accurately identify and handle high-risk cases, rationally allocate the time and effort of business personnel in handling cases, and promptly reduce risks and losses.

A method for model training, the method comprises: obtaining a data sample for a target machine learning model to be trained; and training the target machine learning model by performing an iterative process comprising: generating, using the target machine learning model, a first likelihood of a predicted response token based on a current token sequence, generating an advantage score for a predicted model response with respect to the current token sequence based on the sample model response, the predicted model response being updated iteratively to comprise the predicted response token, determining a first loss value of a first loss function for the target machine learning model based on the generated likelihood and the advantage score, and updating the target machine learning model based on the first loss value.

The invention discloses a recommendation method based on interpretable generalized logic transformation matrix decomposition. The recommendation method comprises the following steps: converting an original scoring matrix into normal distribution data through a generalized logic transformation function; constructing indexes based on similarity and indexes based on ranking; calculating probability distribution and expected scores of scores of the recommended items by the similar users, and generating interpretability indexes in combination with the similarity indexes; integrating the interpretability index into a matrix decomposition objective function for optimization; solving a user feature matrix and a project feature matrix through an alternating least square method or stochastic gradient descent; calculating a prediction score and mapping the prediction score back to an original score interval through generalized logic inverse transformation; generating a recommendation list; the method provided by the invention has wider applicability and higher performance in practical application.

The invention discloses a random sampling consistency GNSS (Global Navigation Satellite System) positioning method fusing random forest prior information, which comprises the following steps of: firstly, carrying out random forest classifier training according to multi-dimensional feature information and a label of each GNSS observation value; secondly, predicting each observation value of the current epoch by using the trained random forest classifier to obtain a probability score of normal observation; then, in a random sampling consistency iteration process, sorting all minimum subsets according to the sum of probability scores of observation values in each minimum subset, and preferentially selecting a subset with a high probability to carry out model parameter estimation; and then, based on a kernel density estimation method, probability density evaluation is carried out on all candidate positioning results, and a position with the maximum probability density is selected as a final positioning solution, so that the stability and credibility of the results are enhanced. According to the method, the positioning precision, the stability and the anti-interference capability of the GNSS system in the urban complex environment are effectively improved.