336 results about "Model selection" patented technology

An apparatus comprising an interface and a processor. The interface may be configured to receive sensor data from a plurality of sensors. The processor may be configured to detect an intruder in response to an analysis of the sensor data, activate an AI model in response to detecting the intruder and generate a countermeasure signal in response to a countermeasure selection by the AI model. The AI model may be configured to analyze the sensor data, compare the sensor data of the intruder to a database of pests to perform a classification of the intruder, determine the countermeasure selection in response to the classification of the intruder as a selected pest, monitor the sensor data for an outcome of the countermeasure selection, and generate a text description of the classification of the selected pest, the countermeasure selection and the countermeasure outcome.

Aspects of the present disclosure provide systems, methods, and computer-readable storage media that support intelligent model selection for style-specific digital content generation. For example, a system that provides a digital content generation service may include a trained style detection model may receive reference digital content items from a user and extract a user style embedding that represents a style preference of the user. In some implementations, the reference digital content items may include text documents or images provided or selected by the user. The system may compare the user style embedding to a plurality of model style embeddings that each correspond to a respective generative artificial intelligence (AI) model to generate a ranked list of generative AI models. The system may access one or more highest ranked generative AI models from the ranked list to generate novel digital content based on a prompt from the user.

Methods and systems are described for AI and / or ML model selection in a telecommunications network. One example embodiment includes obtaining a first set of one or more criteria for selecting between at least two AI / ML models; performing at least one measurement procedure to obtain one or more measurements; and using the obtained one or more measurements and the first set of one or more criteria to select between the at least two AI / ML models. A measurement procedure can comprise e.g. CSI, radio link procedure (RLP), positioning measurement, measurement related to cell change procedure etc. Certain described embodiments enhance measurement performance of the measurement (e.g., CQI) and correspondingly improve the outcome / performance of the procedure (e.g., data scheduling) using the measurement. This in turn reduces the overall processing in the UE, frees up at least part of the memory resources and reduces the UE power consumption.

The invention discloses an aggregate random distribution concrete PFC (Power Factor Correction) simulation method, which comprises the following steps: a model selection step: determining a parallel bonding model as a mechanical model for concrete material simulation; a cement mortar matrix generation step of generating a cement mortar matrix by using particles in a set size range in a set space range based on the model, and setting pore density, particle density and contact model parameters; a coarse aggregate random distribution simulation step: randomly covering and grouping original particles in the matrix in a grouping and range mode; a concrete numerical model construction step: determining particle classification so as to distinguish cement mortar and coarse aggregate; and a model reliability verification step: comparing indoor test results through a uniaxial compression test. According to the method, the simulation accuracy is improved, the randomness of aggregate is reduced, the model reliability is ensured, the matrix simulation refinement degree is improved, and a method is provided for concrete modeling and performance research.

The invention provides a fault diagnosis method, diagnosis device and fault diagnosis equipment, and relates to the technical field of fault diagnosis, and the fault diagnosis method comprises the steps: obtaining test firmware corresponding to the model of a to-be-measured force plate in response to the model selection of the to-be-measured force plate, and displaying a test interface corresponding to the to-be-measured force plate; and in response to a detection operation for the to-be-measured force plate, running the test firmware, and determining the working state of the to-be-measured force plate. According to the method, efficient fault diagnosis is realized through the automatic test process and the user interaction interface, the efficiency and accuracy can be improved, and manual dependence is reduced.

The invention discloses a multi-dimensional influence index evolutionary game evaluation method, system and device for offshore wind power delivery system planning and a medium. The method comprises the steps that an offshore wind power delivery system planning multi-dimensional influence index system is constructed; constructing a subjective and objective weight calculation method, performing subjective weighting on the dimension influence index, and performing objective weighting on the information of the dimension influence index data sample; various subjective and objective weighting mechanisms are modeled into a strategy population according to multi-weighting criterion decision analysis, a strategy proportion is dynamically adjusted in an evaluation task by constructing a fitness function and copying a dynamic evolution mechanism, a weighting optimization model is constructed, and an index combination weight is obtained. The finally obtained stable comprehensive weight vector can significantly improve the scientificity, robustness and engineering applicability of an evaluation result, and provides support with higher decision guidance value for offshore wind power delivery system planning and technical model selection.

The invention discloses an education question answering model routing method based on deep learning and a knowledge graph, and aims to solve the problems that the existing multi-model selection depends on text similarity, lacks knowledge points and first repair constraints, is easy to route by mistake and generates illusion. The method comprises the following steps: generating problem sub-graphs on an educational knowledge graph, carrying out relation perception graph neural network coding, constructing a model capability graph containing expression distribution and service attributes, forming masked cost under the constraints of types, privacy, time delay, cost and first repair coverage, and obtaining soft assignment by adopting unbalanced optimal transmission with entropy regularization; and interpretable routing is completed in combination with the meta-path attention subjected to interpretation consistency regularization training, so that the technical effects of reducing wrong routing and illusion rate, improving preferred correct routing rate and outputting meta-path interpretation are achieved.

The invention belongs to the technical field of data processing, and particularly discloses a computer vision model service scenarized configuration generation method, which comprises the following steps: constructing a basic information database comprising a plurality of computer vision models, the database being used for storing model identification information and feature parameters; extracting structured features of the scene according to a scene analysis engine, wherein the structured features are used for describing operation conditions of the scene; calculating a model matching score according to the structured features and the feature parameters, and obtaining a recommendation model through a multi-condition weight matching function; executing a model selection strategy based on the matching score, and obtaining a model matched with the scene; and generating a configuration file conforming to a predetermined specification according to the recommendation model. The objective of the invention is to solve the problems of low efficiency and error proneness caused by the fact that a computer vision model in the prior art depends on a manual selection model and cannot quickly select and adapt to an optimal model in diversified scenes.

Systems and methods for handling Artificial Intelligence (AI) models for Human Interface Devices (HIDs). In some embodiments, an HID may include a processor; and a memory coupled to the processor, the memory having program instructions stored thereon that, upon execution, cause the HID to: receive an AI model selection from the user; and in response to the AI model selection, load a corresponding AI model configured to receive raw user input and to produce processed user input.

Embodiments predict a target variable for accounts receivable using a machine learning model. For a first customer, embodiments receive a plurality of trained ML models corresponding to the target variable, the plurality of trained ML models trained using the historical data and comprising a first trained model having no grace period for the target variable and two or more grace period trained models, each grace period trained model having different grace periods for the target variable. Embodiments determine a Matthews' Correlation Coefficient (“MCC”) for the first trained model. When the MCC for the first trained model is low, embodiments determine the MCC for each of the grace period trained models, and when one or more MCCs for each of the grace period trained models is higher than the MCC for the first trained model, embodiments select the corresponding grace period trained model having a highest MCC.

A method comprises receiving a current dataset for a current time window from at least one sensor in a wellbore created in a subsurface formation, wherein the current dataset comprises values of a number of current features of the subsurface formation at a spatial location in the wellbore. The method includes selecting at least one previous time window from a number of previous time windows that includes a previously cached dataset that was detected by the at least one sensor or a different sensor in the wellbore and that spatially overlaps with the spatial location for the current dataset. The method includes merging the current dataset with the previously cached dataset to create a merged dataset. The method includes selecting a machine learning model from a plurality of machine learning models for the spatial location in the wellbore based on the merged dataset.

The invention provides a low-power-consumption cooperative sewer water body integrated monitoring method. The method comprises the following steps: carrying out model selection and deployment on sensing node hardware of an adaptive scene; a low-power-consumption internet of things network topology and power supply design; a time-sharing triggering acquisition strategy of the multi-modal data is carried out; node-level environment self-adaptive data calibration is carried out; node-level flow velocity-stink association anomaly recognition is carried out; carrying out network-level data fusion and validity verification; collaborative traceability, diffusion simulation and linkage control are carried out. According to the invention, the problems of data distortion, difficulty in source tracing of pollution sources, short equipment endurance and response lag caused by independent collection of flow and stink and environmental interference in traditional sewer monitoring are solved; the integrated effect of ultrasonic Doppler and electronic nose cooperative monitoring, environment self-adaption accurate calibration, odor source rapid positioning and diffusion simulation and low-power-consumption and long-endurance linkage control is achieved.

A stratum reinforcement modeling and analysis method for a shield tunnel crossing an existing structure is provided. Geological information and an engineering structure condition of the existing structure are acquired. Physical parameters of different soil mass layers are determined based on a laboratory test. A three-dimensional (3D) model is constructed. Mesh generation is performed on the 3D model defining a constraint for the 3D model. Model parameters are selected for the 3D model. A soil mass constitutive model is selected for the 3D model. Synchronous grouting simulation, construction load simulation and shield tunneling construction simulation are performed on the 3D model. Stratum parameters are changed to achieve stratum reinforcement modeling. Data analysis is performed by adopting different reinforcement schemes to achieve stratum reinforcement modeling and numerical analysis of the process of the shield tunnel crossing the existing structure.

An adaptive artificial intelligence (AI) control system receives a prompt for an AI system from a user interface component of a software application on a mobile device. The adaptive AI control system determines a current battery level of the mobile device using a battery level monitoring component. The adaptive AI control system then selects a generative AI model of the AI system to use to generate a response for the prompt based on the current battery level using a model selection component. The generative AI model that is selected is one of a plurality of different generative AI models of the AI system which are capable of processing the prompt, each of the plurality of generative AI models having a different level of complexity.

The invention discloses a watershed water level missing data interpolation method, which belongs to the technical field of water level data interpolation, and comprises the following steps: (1) data preprocessing: collecting hourly water level data and hydrological data of a plurality of stations of a target river reach, carrying out integrity and consistency inspection, and evaluating the data quality grade of each station; (2) model construction: respectively constructing an interpolation model based on a time sequence, an interpolation model based on spatial features and an interpolation model based on an interpolation algorithm based on the data quality grade; (3) sub-scene model selection: formulating an interpolation model selection strategy according to a data missing type and missing duration, and performing data interpolation; and (4) verifying interpolation precision through evaluation indexes, and outputting repair data meeting requirements. According to the method, the problems of single adaptive scene and insufficient precision in the prior art are solved, high-precision and stable interpolation of the water level data under multiple missing scenes is realized, and the requirements of hydrological analysis, flood control scheduling and shipping safety are met.

Techniques for generating augmented prompts are disclosed herein. Augmented prompts and / or guardrails for augmenting prompts are identified and / or generated. Augmented prompts intended to secure output generated by a generative model resulting from the augmented prompts are scored for efficacy. A risk classifier and a rules-based dictionary for augmenting prompts according to a risk class of an initial prompt are used to generate training data. The training data is used to train and / or fine-tune an error-to-prompt model. Augmented prompts and / or efficacy scores for the augmented prompts are used for feedback-based optimization of the error-to-prompt model. The error-to-prompt model selects and / or generates prompt augmentations, such as guardrail phrases, edits, deletions, or the like that secure output generated by the augmented prompt.

Provided are systems for ensemble learning with machine learning models that include a processor to receive a training dataset of a plurality of data instances, wherein each data instance comprises a time series of data points, add an amount of time delay to one or more data instances to provide an augmented training dataset, select a first plurality of supervised machine learning models, select a second plurality of unsupervised machine learning models, train the first plurality of supervised machine learning models and the second plurality of unsupervised machine learning models based on the augmented training dataset, generate an ensemble machine learning model based on outputs of the supervised machine learning models and unsupervised machine learning models, and generate a runtime output of the ensemble machine learning model based on a runtime input to the ensemble machine learning model. Methods and computer program products are also provided.

A system for Time Series Generation (TSG) model selection. The system is configured to perform a method including: receiving a user prompt, the user prompt comprising a user input and a user-provided TSG dataset; providing the user input as input to a first machine learning model, to determine the user prompt as a TSG query; providing the user input and the user- provided TSG dataset as input to a second machine learning model, to select at least one shortlisted TSG model from a TSG database; and using the first machine learning model, providing the at least one shortlisted TSG model as a response to the TSG query.

The invention provides an attribute error binding attack and evaluation method for an identity preserving model, which comprises the following steps of: acquiring an open source unadapted content data set which comprises a plurality of cue words, labeling the cue words along four harmful dimensions, decomposing the cue words into three semantic components, constructing a sensitive word data set based on the plurality of cue words subjected to labeling and component decomposition; sensitive words are extracted to serve as input cue words, candidate cue words are generated by selecting an error binding strategy through an LLM model, and effective cue words are integrated into an error binding instruction evaluation set; obtaining an identity reference image, inputting the identity reference image into a diffusion model of the identity maintenance model, and generating a generated image which maintains character identity features and accords with cue word description; and performing attribute binding security evaluation on the generated image to obtain an attribute binding security score. The technical problem that the attack risk based on'attribute wrong binding 'cannot be effectively evaluated in the prior art is solved.

The invention discloses a MaaS platform construction method and system based on evaluation-driven closed-loop optimization, and the method comprises the steps: enabling a model square to serve as a unified entrance of a model, and providing rich model resources for a user to select; the computing platform provides model development, training and deployment for a user; evaluating the performance of the model; through feedback, close cooperation among the model square, the calculation platform and the model evaluation is realized, and an automatic optimization closed loop is formed. According to the closed-loop optimization mechanism based on evaluation driving, close cooperation among the model square, the calculation platform and the model evaluation module is achieved through feedback of the model evaluation module, an automatic optimization closed loop is formed, and the problem that an existing platform cannot achieve the automatic process from model selection, evaluation to retraining is solved.

A method includes obtaining training data for a retrieval-augmented generation (RAG) architecture having retriever and generative models. The retriever model is configured to identify information chunks relevant to input queries, and the generative model is configured to generate outputs based on the information chunks and the input queries. The method also includes generating a prompt for the generative model and generating multiple sets of queries for the retriever model. Each query in the multiple sets of queries is configured to cause the retriever model to select a set of information chunks associated with the prompt. The method further includes generating multiple responses to the prompt using the generative model and the sets of information chunks and determining rewards associated with the RAG architecture based on the responses. In addition, the method includes training the generative model based on the training data and the rewards to produce an updated RAG architecture.

The invention discloses an osmotic pressure analysis method and device fusing model optimization and physical inversion and a medium, and the method comprises the steps: calculating correlation coefficients of a specified reservoir water level and osmotic pressure under different time lags based on cross-correlation function analysis, and recognizing a differential nonlinear feature corresponding to the maximum correlation coefficient; aiming at the differentiated nonlinear characteristics of different dam types, constructing a plurality of corresponding diagnosis prediction models in parallel, and establishing a model selection decision rule to select an optimal diagnosis model; according to delay parameters obtained through cross-correlation analysis, a permeability coefficient is reversely deduced through a pore medium heat transfer diffusion theoretical formula; according to the regression slope and the theoretical value attenuation factor of the optimal model, a permeability coefficient is reversely deduced in combination with a correction diffusion equation, credibility evaluation and weighted fusion are performed on the permeability coefficient, and a comprehensive permeability coefficient is calculated; and diagnosing the soil type and the permeability characteristic grade. According to the method, the propagation delay characteristic of reservoir water level change in the pore medium can be accurately quantified, and the inversion precision of the permeability coefficient and the engineering applicability are improved.

Model selection is disclosed. Features used as inputs to models are scored in terms of importance and health. The importance and health scores are combined in order to generate a model score for each model. The model with a score above a threshold score is selected and deployed.

The invention provides a real-time time sequence prediction system and method based on a dynamic weight hybrid model, and belongs to the technical field of time sequence prediction and machine learning. The prediction system comprises a data generation module, a real-time data caching module, a model initialization module, a prediction model selection module, a single model prediction module, a mixed model prediction module, a result display module, an error calculation module, a dynamic weight optimization module and a reset module. According to the method, a dynamic weight fusion strategy is adopted, and contribution weights of ARIMA and LSTM models are dynamically adjusted according to real-time data characteristics (such as data stability, non-linear degree and fluctuation amplitude). A dynamic weight mechanism solves the problem of'one-cut 'of a fixed weight hybrid model, so that the model can maintain the optimal performance in a linear stable scene (such as a new energy output stable time period) and a nonlinear fluctuation scene (such as an extreme weather time period), and the generalization ability of prediction is remarkably improved.

The invention relates to the technical field of animal genetic breeding, and provides a double-character genome selection platform based on local genetic correlation (LGC). The platform is composed of a phenotype data processing module, a genotype data processing module and a genome selection module (comprising a model selection sub-module and a parameter selection sub-module). The platform performs quality control, data normality test and correction on phenotypic data; performing quality control, sequence alignment, variation detection and genotype filling on the genotype data; and identifying whole genome local genetic correlation, checking the significance of LGC, selecting a model based on a whole genome LGC estimation result of a character pair and a corresponding P value, setting a corresponding model parameter threshold, and realizing section weighted double-character genome selection analysis. According to the method, the genetic evaluation accuracy of complex characters, especially low heritability characters, can be effectively improved, and an efficient and stable genome selection tool is provided for livestock and poultry breeding.

The application discloses a content generation and interactive dialogue method and system based on AI artificial intelligence, which models semantic label sequences in multiple rounds of conversation, identifies semantic derailment points in behavior evolution trajectory templates and triggers an intent clarification mechanism; constructs a semantic tension map based on user feedback, forms a tension closed cluster and generates multi-path response candidate segments; uses a misleading chain cross-analysis unit to identify high-risk response chains associated with historical negative feedback, and eliminates and shields content; and selects the optimal response segment through a context break span model and dynamically updates the behavior evolution model. The application can significantly improve the semantic stability, consistency and safety of the dialogue system, reduce semantic drift, misleading answers and illusion risks, and has high reliability and technical advancement. The system scheme can realize the modularization of the method functions and is suitable for application in intelligent customer service, general large models, decision support systems and other scenes.

This application provides a method for adjusting the quantization coefficients of a frame and a method for encoding a video stream. The method includes: acquiring a target image block corresponding to a target frame; determining model selection parameters corresponding to the target image block, wherein the model selection parameters include at least one of the following: block size parameters, and other parameters; retrieving a target model corresponding to the model selection parameters from multiple candidate models; determining input parameters corresponding to the target model, wherein the input parameters include target block coefficients of the target image block, target quantization parameters, and initial quantization coefficients, and the target block coefficients include target amplitude coefficients and target sign coefficients; inputting the input parameters into the target model to obtain a predicted adjustment value for the target image block; and adjusting the initial quantization coefficients based on the predicted adjustment value to obtain the target quantization coefficients. This solves the technical problem in related technologies of balancing compression efficiency and computational complexity.

This invention belongs to the technical field of multi-scale particle transport simulation, and discloses a kinetic solution method and system for multi-scale particle transport simulation. The method includes: a particle kinetic model based on discrete velocity form; selecting discrete points in velocity space; calculating the distribution function along the direction of the discrete points in velocity space at the center of the grid according to a compact scheme; accumulating the contribution values ​​of the distribution function to macroscopic physical quantities; scanning the downstream grid sequentially based on the direction of the discrete points in velocity space; changing the discrete points in velocity space and repeating the operation until the contribution values ​​of the grid at all discrete points in velocity space are accumulated, and obtaining the final macroscopic physical quantities; outputting the calculation results of macroscopic physical quantities when the convergence condition is met. This invention discloses a kinetic solution method and system for multi-scale particle transport simulation that balances computational memory, accuracy, and efficiency, and is particularly suitable for multi-scale particle transport simulation with a large number of discrete points in velocity space and strong heterogeneity.

Aspects of the disclosure are directed to methods, systems, and computer readable media for adaptation with self-evaluation to improve selective prediction in large language models (LLMs), generally referred to as ASPIRE. ASPIRE includes training LLMs on a portion of training data from a question answering task to learn self-evaluation, e.g., learn to distinguish whether a generated answer is correct or not. ASPIRE further includes a selection score that combines a likelihood of that generated answer is correct with a self-evaluation score for selective prediction. ASPIRE demonstrates improved selective prediction performance with less computational cost.

The invention discloses a targeted design and optimization method for an organic pollutant fluorescence sensing array and a signal acquisition device. The overall information of the array is obtained in a high-throughput manner in a fluorescence imaging mode; based on an inner filter effect sensing mechanism, the model selection of a sensing material and an optical device is carried out by taking the absorption spectrum characteristic of a target organic detection object as a starting point, and sensing optimization is carried out based on the controllability of weak interaction between the sensing material and the target detection object. The use process comprises two aspects of sample library construction and modeling and detection application. The method comprises the steps of sensing material selection, model selection of an excitation light source and a signal acquisition device, light path arrangement design and sensing performance optimization. The detection device disclosed by the invention has proper detection precision and portability, can replace a precise fluorescence spectrophotometer, and realizes low-cost, high-throughput and real-time field detection of a water body.