47 results about "Statistical learning" patented technology

The invention relates to a rock slope support model construction scheme generation method and device, equipment and a medium. According to the method, a three-dimensional geological model fusing geological information and rock mass parameters is constructed, an initial damage field is obtained in combination with micro-seismic monitoring data and statistical learning inversion, and then a constitutive model capable of reflecting the damage and plastic coupling evolution law is established and calibrated; the model is used for dynamically predicting a spatio-temporal evolution path of a potential slip plane in the excavation process, the supporting opportunity and position are accurately judged based on the stress and damage state in the path, a spatio-temporal sequence scheme is generated, and finally optimal supporting parameters are solved through the multi-objective optimization model. And finally, a set of dynamic support construction scheme capable of actively controlling damage development and giving consideration to safety and economical efficiency is integrated and output, technical spanning from passive reinforcement to active intervention and from static design to dynamic optimization is achieved, and the accuracy and reliability of slope support are effectively improved.

The invention discloses a few-sample inductance core surface defect detection method based on model interaction, and belongs to the technical field of machine vision and industrial defect detection. The method comprises the following steps: S1, data acquisition and image preprocessing are carried out, and normal samples, labeled samples and unlabeled samples are constructed; s2, constructing an unsupervised statistical model based on statistical learning; s3, constructing a supervised semantic segmentation model; s4, inputting an inductance core image to be detected into the unsupervised statistical model and the supervised semantic segmentation model at the same time for processing, and generating a segmentation result; s5, detecting result differences are quantified; s6, performing parameter updating on the unsupervised statistical model; s7, generating a pseudo label based on an unsupervised statistical model; s8, carrying out weight updating on the supervised semantic segmentation model; and S9, inputting the processed to-be-detected images into the updated unsupervised statistical model and supervised semantic segmentation model in batches for detection to obtain a detection result, and analyzing and calculating system performance indexes.

Systems, methods, and apparatus for automatic signal detection in a radio-frequency (RF) environment are disclosed. At least one node device is in a fixed nodal network. The at least one node device is operable to measure and learn the RF environment in a predetermined period based on statistical learning techniques, thereby creating learning data. The at least one node device is operable to create a spectrum map based on the learning data. The at least one node device is operable to calculate a power distribution by frequency of the RF environment in real time or near real time, including a first derivative and a second derivative of fast Fourier transform (FFT) data of the RF environment. The at least one node device is operable to identify at least one signal based on the first derivative and the second derivative of FFT data.

The invention relates to the technical field of biostatistics, and discloses a noise language frequency hearing impairment prediction system based on a GEE model and genetic characteristics, and the system comprises a data collection module which collects follow-up visit data of a worker; the data processing module is used for generating a standardized modeling data set; the feature screening module is used for screening features through a statistical learning method to obtain a key feature subset; the time-varying interaction module is used for constructing a generalized estimation equation model and outputting regression coefficient estimation; the risk prediction module outputs the risk probability and the risk layering result of the individual; the decision support module is used for generating hearing protection suggestions of the individuals; according to the method, the generalized estimation equation model is constructed to process follow-up data of occupational noise exposure workers, so that the accuracy and the stability of language frequency hearing loss risk prediction are improved, a basis is provided for formulating a personalized hearing protection scheme and recommending a proper hearing protection device, and the method is suitable for popularization and application. And the transformation of occupational hearing loss from passive treatment to active prevention is facilitated.

The invention discloses a user self-filling washing requirement and standard process difference verification method, and particularly relates to the technical field of life, and the method comprises the steps: obtaining a corresponding standard process parameter vector from a clothes feature vector through a standard process retrieval algorithm based on a process knowledge base, constructing a knowledge base secondary retrieval index, and traversing the process knowledge base; and obtaining a candidate set meeting the constraint, screening out a record with the highest score, and obtaining a recommendation process and a security boundary. The method is realized based on a standard parameter vector, a rule engine and a statistical learning method, can be conveniently integrated into an existing internet clothes washing platform or a traditional clothes washing management system, can be flexibly expanded according to standard process libraries of different enterprises, and is easy to integrate and expand.

Systems, methods, and apparatus for automatic signal detection in a radio-frequency (RF) environment are disclosed. At least one node device is in a fixed nodal network. The at least one node device is operable to measure and learn the RF environment in a predetermined period based on statistical learning techniques, thereby creating learning data. The at least one node device is operable to create a spectrum map based on the learning data. The at least one node device is operable to calculate a power distribution by frequency of the RF environment in real time or near real time, including a first derivative and a second derivative of fast Fourier transform (FFT) data of the RF environment. The at least one node device is operable to identify at least one signal based on the first derivative and the second derivative of FFT data.

The invention provides an intelligent detection system and method for hidden cracks and fragments of a silicon wafer, and the system comprises an imaging unit, an image processing unit and a control and communication unit, and is used for obtaining a surface image of the silicon wafer located on a conveying belt; the image processing unit is in communication connection with the imaging unit and is used for receiving and processing the surface image, the image processing unit comprises a processor, and the processor is configured to automatically generate an effective detection area excluding an edge area and an electrode area in the surface image based on pre-stored silicon wafer size information and electrode distribution information; the processor is used for segmenting the image in the effective detection area based on a gray threshold range determined through statistical learning, extracting a suspected defect area, performing feature extraction on the suspected defect area, and performing defect classification and judgment according to preset defect feature parameters; and the control and communication unit is connected with the image processing unit, so that efficient and accurate subfissure and fragment detection is realized on the premise of not changing the structure of a production line.

The application relates to the technical field of hydraulic engineering, and discloses a soil and rock dam dam-break probability prediction method, which comprises the following steps: acquiring basic cause variables and state conversion variables leading to soil and rock dam dam-break, adopting a fixed threshold method to perform discretization processing, and generating discretization standards of the variables; a one-dimensional dam-break flow simulation model is constructed; based on the discretization standards, a data set covering multiple grades of flood scenarios is generated; a three-layer Bayesian network model is constructed; the data set is input into the three-layer Bayesian network model for training; a Bayesian smoothing method is adopted to calculate the conditional probability of each node in the three-layer Bayesian network model, so as to correct model parameters; finally, the trained three-layer Bayesian network model is input into the variable data of a to-be-tested soil and rock dam, so as to realize dam-break probability prediction and risk grade evaluation; through fusion of a physical mechanism and statistical learning, the dam-break prediction probability is quantified and improved, and high-precision dam-break risk prediction is realized.

The invention belongs to the technical field of power system dynamic safety analysis and renewable energy source grid connection, and provides a method for determining the permeation limit of high renewable energy sources under multi-dimensional stability constraint. The method takes physical mechanism depth modeling-multi-dimensional stability coupling theory-intelligent data driven screening as a core, constructs a quantitative correlation model of frequency stability, transient stability and small disturbance stability through rigorous theoretical derivation, proposes a global-region double-layer dynamic inertial constraint mechanism, and combines an AI-driven critical scene screening and clustering method to obtain a critical scene clustering algorithm. And finally determining the safe RE penetration limit. According to the method, the limitation of traditional single-dimension evaluation and static constraint is broken through, all theoretical derivation is based on the dynamic characteristics and the statistical learning principle of the power system, moderate data verification is assisted, and a complete theoretical system and an engineering tool are provided for RE safety integration of the low-inertia power grid.

1. The name of the design product: college student focused habit forming graphical user interface of mobile phone. 2. The use of the design product: for running programs and displaying graphical user interfaces. 3. The design points of the design product: in the graphical user interface. 4. The picture or photo that best indicates the design points: front view. 5. The use of the graphical user interface: to improve the learning efficiency and self-discipline of college students, by recording and counting the learning time, the focused results can be seen directly, and the formation of learning habits is promoted. 6. The change state of the graphical user interface: the front view is the initial interface of the college student focused habit forming application, clicking the "concentrate clock" icon below the front view enters change state figure 1, clicking the round icon at the upper right corner of change state figure 1 enters change state figure 2. 7. Other circumstances that need to be explained: "XXX" in the view represents replaceable text and / or numbers and / or letters and / or symbols, and the gray block part in the view is the content screen.

Systems, methods and apparatus are disclosed for automatic signal detection in an RF environment. An apparatus comprises at least one receiver and at least one processor coupled with at least one memory. The apparatus is at the edge of a communication network. The apparatus sweeps and learns the RF environment in a predetermined period based on statistical learning techniques, thereby creating learning data. The apparatus forms a knowledge map based on the learning data, scrubs a real-time spectral sweep against the knowledge map, and creates impressions on the RF environment based on a machine learning algorithm. The apparatus is operable to detect at least one signal in the RF environment.

The invention relates to the technical field of hydraulic engineering, and discloses an earth and rockfill dam break probability prediction method, which comprises the following steps of: obtaining a foundation dependent variable and a state conversion variable which cause earth and rockfill dam break, performing discretization processing by adopting a fixed threshold method, and generating a discretization standard of each variable; constructing a one-dimensional dam break water flow simulation model, and generating a data set covering a multi-level flood scene based on a discretization standard; constructing a three-layer Bayesian network model; inputting the data set into a three-layer Bayesian network model for training, calculating the conditional probability of each node in the three-layer Bayesian network model by adopting a Bayesian smoothing method so as to correct model parameters, and finally, after the trained three-layer Bayesian network model is generated, inputting each variable data of the earth and rockfill dam to be tested, therefore, dam break probability prediction and risk level evaluation are realized. By fusing a physical mechanism and statistical learning, the dam break prediction probability is quantified and improved, and high-precision dam break risk prediction is realized.

A method includes obtaining information representing a current state of communication with a user equipment (UE) performed using one or more beams. The method also includes comparing the information to statistical historical state information to determine one or more best next narrow beam candidates. The method further includes performing a beam search using the one or more best next narrow beam candidates in order to select a next narrow beam. The method also includes communicating with the UE using the selected next narrow beam.

A method for using a Gaussian Process-based algorithm to approximate an optimal stopping of a time series that corresponds to a sequence of events is provided. The method includes: receiving information that relates to an event sequence; estimating, based on the received information, a first potential reward that is obtained by stopping the event sequence at a first time, and a set of respective second potential rewards that are obtained by stopping the event sequence at corresponding times; and determining, based on the estimated first and second potential rewards, an optimal time for stopping the event sequence. The event sequence may include a numerical sequence that is modeled as a statistical learning method via a Gaussian Process (GP) function and / or a deep GP function that indicates a probability density distribution of the items in the numerical sequence over a predetermined time interval.

The application discloses a user travel evaluation sentiment analysis method and device. The method comprises the following steps: preprocessing a user travel evaluation text to obtain a feature matrix; inputting the feature matrix into a trained random forest sub-model to output a first sentiment probability; inputting the feature matrix into a trained LSTM model with attention mechanism to output a second sentiment probability; fusing the first and second sentiment probabilities by weighting to obtain a third sentiment probability; inputting the feature matrix, the first sentiment probability and the second sentiment probability into a generator to output a sentiment analysis report; determining the authenticity of the sentiment analysis report by using a discriminator combined with a confidence threshold; obtaining a fourth sentiment probability; and fusing the third and fourth sentiment probabilities to obtain a final sentiment probability. The method disclosed in the application fuses a hybrid architecture of statistical learning and deep learning, and introduces an adversarial learning mechanism to realize dynamic optimization of sentiment analysis results, so that the sentiment judgment precision is improved through multi-level model cooperation.

The invention belongs to the field of image processing, and discloses an underwater image quality evaluation method and system based on visual contrastive analysis and brightness-chrominance statistical learning, which enable non-reference quality evaluation to have class reference capability by generating a pseudo original image corresponding to a distorted image to be detected, and remarkably improve the effectiveness of quality difference measurement. By comparing and analyzing the pseudo original image and the distorted image to obtain the first perception quality difference feature, the real perception change of underwater distortion can be described, and the accuracy of perception difference measurement is improved. Second perception quality difference features are calculated based on the similarity of the two, so that the evaluation model can comprehensively consider the structural consistency and the detail degradation degree, and the final quality judgment is more fit for the visual perception rule. According to the method, the accuracy and robustness of no-reference underwater image quality evaluation are greatly improved, and the method is suitable for more engineering application scenes.

The invention relates to a method for predicting the content of elements in coal based on LIBS spectral feature optimization and machine learning, which comprises the following steps of: converting a preprocessed spectrum into a two-dimensional spectral matrix, matching a peak wavelength point of the two-dimensional spectral matrix with an NIST standard atomic spectrum database, identifying spectral lines belonging to target elements, and determining the content of the elements in the target elements according to the spectral lines. A characteristic wavelength set stably appearing in all coal sample spectrums is screened out to serve as candidate characteristics; further screening out a core characteristic wavelength subset which is most relevant to the content of the target element and has the minimum redundancy by applying a plurality of characteristic selection methods; and then selecting an optimal model combination by adopting a plurality of machine learning regression algorithms. According to the method, through a two-stage screening strategy combining physical spectral line identification (NIST matching) and statistical learning feature selection, irrelevant and redundant spectral information is eliminated to the greatest extent, anti-interference core features directly related to the target element content are reserved, and the quality of model input data is improved from the source.

The embodiment of the application provides an intelligent ore-prospecting method and device based on multi-modal data and a control theory model, which comprises the following steps: obtaining a multi-modal data body by fusing and aligning multi-source heterogeneous data of a target mining area; extracting key geological slow variables from the data body as system order parameters based on the Haken slave principle and sparse statistical learning, and classifying the key geological slow variables into resistance and driving force control variables for gridding; then, applying the Thom cusp catastrophe theory to calculate a mutation discriminant value in the whole area; finally, intelligently delineating and optimizing an ore-prospecting target area through anomaly area identification, uncertainty analysis and Bayesian updating driven by real drilling data, and a high-confidence exploration target. The application can improve the accuracy and efficiency of deep ore prospecting.

The invention relates to an AI-based medical-care integrated data management method and system, electronic equipment and a storage medium. The method comprises the following steps: acquiring multi-source heterogeneous data of a target object; performing data cleaning on the multi-source heterogeneous data based on a hierarchical cleaning strategy to obtain target data; performing feature alignment on the feature vectors in the target data based on a metric learning algorithm to generate a fused feature vector; reasoning and evaluating the fusion feature vector based on a hybrid expert model to obtain an auxiliary diagnosis result; wherein the hybrid expert model integrates a traditional Chinese medicine rule engine and a western medicine statistical learning engine. Therefore, the quality problem of the multi-source heterogeneous data is solved in a targeted manner through a hierarchical cleaning strategy, so that the interference of invalid data is reduced, and the accuracy and integrity of the data are ensured; based on a feature alignment algorithm of metric learning, a cross-domain mapping relation can be established, multi-dimensional data form complementation, a fusion feature vector is finally generated, and the data representation capability is enhanced.

The invention relates to the technical field of ecological hydrological modeling and artificial intelligence crossing, in particular to a multi-source satellite data driven irrigation area ecological hydrological basic model construction method, which comprises the following steps: acquiring multi-source satellite data and ecological hydrological related data of a target irrigation area; preprocessing the multi-source satellite data and the ecological hydrological related data, and interpolating missing measurement data to generate complete ecological hydrological data; according to the method, time sequence satellite general characterization generated by multi-source satellite data and physical variable embedding of ecological hydrological data are fused, primary ecological hydrological general characterization is generated, an irrigation area ecological hydrological basic model is constructed in combination with a self-supervision task, unified ecological hydrological general characterization is output, and ecological hydrological variable estimation is carried out. Therefore, the problems that in related technologies, due to dependence on statistical learning and lack of physical constraints, result time sequences are discontinuous or do not conform to physical laws, multi-source data fusion is insufficient, and high-precision estimation of ecological hydrological elements is difficult to achieve are solved.

This invention discloses a few-sample inductor core surface defect detection method based on model interaction, belonging to the fields of machine vision and industrial defect detection technology. The method includes the following steps: S1 Data acquisition and image preprocessing, constructing normal samples, labeled samples, and unlabeled samples; S2 Constructing an unsupervised statistical model based on statistical learning; S3 Constructing a supervised semantic segmentation model; S4 Simultaneously inputting the inductor core image to be detected into both the unsupervised statistical model and the supervised semantic segmentation model for processing, generating segmentation results; S5 Quantifying the differences in detection results; S6 Updating the parameters of the unsupervised statistical model; S7 Generating pseudo-labels based on the unsupervised statistical model; S8 Updating the weights of the supervised semantic segmentation model; S9 Using batches of processed images to be detected, inputting them into the updated unsupervised statistical model and the supervised semantic segmentation model for detection, obtaining detection results, and analyzing and calculating system performance indicators.

The application discloses a high-performance file hash calculation method based on double-engine adaptive switching, relates to the technical field of data security, and aims to solve the problems that the existing scheme is mostly heuristic scheduling based on static threshold values or empirical rules, lacks residual time estimation and interval control based on online statistical learning, and lacks real-time detection of residual drift and adaptive scheduling based on evidence; the method comprises the following steps: constructing a high-dimensional vector by mapping observation characteristics, adopting joint modeling of recursive least squares with ridge regularization and quantile regression, providing point estimation and upper and lower bounds of single fragmentation and residual time consumption; fusing the extreme values of residual variance and quantile interval to obtain a robust interval, and using Page-Hinkley / CUSUM for mutation detection to trigger adaptive adjustment of fragmentation size, concurrency and engine type; and before engine switching, serializing internal working vectors, processed bytes and intermediate summaries and performing mirror checking to ensure consistency of cross-engine results.

The application discloses a prediction method of a multifunctional radar random waveform sequence, which regards a radar signal sequence as a multi-layer syntax structure of regularly connected mixed "radar word"-"radar phrase"-"radar sentence". The working modes of the radar are transformed into each other according to fixed transition probabilities, each working mode corresponds to a fixed phrase template library, and a phrase is selected from the phrase template library each time to complete the function of the working mode. The prediction method distinguishes the occurrence probability of the radar word under two different situations of mode and mode, and the radar word at the next moment can be predicted according to the historical sequence and the corresponding working mode. The method of the application trains a probability model in a statistical learning mode, has strong anti-noise capability, solves the problem of the reduction of the prediction accuracy rate caused by the random template in an adaptive output mode, and avoids multiple iterations and multiple loops through algorithm optimization and matrix design, so that the calculation efficiency and the anti-noise capability are improved.

The invention provides a glass reflection removal method based on fuzzy prior distribution and posterior probability maximization, and relates to the technical field of image processing. Based on the prior knowledge that the reflected light has a fuzzy characteristic, convolution processing is performed on the reflected light component and the real scene component by using a filter, and a reflection-removing physical model of the reflected light and the real scene is established; a prior probability model of reflected light and a real scene is obtained through statistical learning, the Bayesian theorem is utilized to convert an image reflection removal problem based on a reflection removal physical model into a posterior probability maximization problem, and an energy minimization equation is deduced; an alternating direction multiplier method is adopted, auxiliary variables are introduced to decompose an energy minimization equation into two sub-problems about a scene and reflected light, and a real scene image after reflection removal is obtained through Fourier transform iteration solution. The method has the advantages of being high in precision, good in generalization and low in deployment cost, and the dual requirements of machine vision scenes for image quality and processing efficiency are effectively met.

The invention relates to a watershed hydrological prediction method and system based on meteorological driving and causal reconstruction, and belongs to the technical field of watershed hydrological prediction. The method comprises the following steps: constructing a causal structure diagram between meteorological driving variables and between meteorological driving variables and rainfall to form an interpretable meteorological and rainfall causal diagram, explicitly converting a functional relationship in a sub-module into a series of structural equations by taking a traditional hydrological model as a construction basis, and uniformly embedding the structural equations into a causal model framework to form a meteorological and rainfall causal diagram; and a unified structural causal graph is constructed through a graph structure merging technology, and the whole process description of a causal path from external climate factors to drainage basin runoff and runoff response is realized. According to the method, a physical mechanism, statistical learning and causal reasoning are fused, the defect that a traditional model lacks explanatory ability and generalization ability when coping with non-stationary meteorological conditions is overcome, and a system modeling method with causal support is provided for flood forecasting, water resource management and extreme climate response simulation.

The invention belongs to the technical field of image data processing, and provides a highly hidden attack recognition and traceability analysis method and system, and the technical scheme is that the method comprises the steps: constructing a time sequence traceability graph, and carrying out the object vectorization and context coding of the time sequence traceability graph; carrying out object vectorization and context coding results on the basis of the time sequence traceability graph, and calculating consistency scores for input edges or local sub-graphs respectively; calculating consistency scores of all the new edges or the local sub-graphs, and combining the consistency scores of all the new edges or the local sub-graphs and a set quantile field to perform hotspot aggregation to obtain candidate aggregation risk scores; fusing the candidate aggregation risk scores into a reward function, performing budgeting causal path search on candidate suspicious sub-graphs through graph strategy reinforcement learning, outputting a minimum full attack hypothesis graph, and executing rule matching and statistical learning on the generated minimum full attack hypothesis graph to obtain an attack type and root cause positioning traceability. The method is suitable for detection and traceability of various novel and unknown attacks.

The invention discloses a method for reducing the number of optical proximity effect correction iterations. According to the method, firstly, statistical learning is carried out on historical correction data, and a bias pre-storage table is established; the table maps different types of graphic segments and their local geometric features (such as line width and pitch) to an optimized feature offset value. When correction is carried out, firstly, corresponding characteristic offset values in the pre-stored table are searched for all segments of the layout to be corrected, primary pre-correction is carried out to obtain a pre-processed layout, and then the pre-processed layout is used as initial input to carry out iterative calculation. By providing a better initial layout, the OPC iteration times are effectively reduced, the operation time is shortened, and the system stability is improved.

The embodiment of the invention provides a training course data processing method, computing equipment, a storage medium and a product. The method comprises the steps of determining a target student user of a to-be-counted learning progress corresponding to a training course and a target course period of current learning of the target student user in the training course, and determining a target course section of the to-be-counted learning progress from a plurality of course sections of the target course period; acquiring learning behavior data of a target student user for a target course section in at least one course period of the training course, determining a target course section template corresponding to the target course section from a plurality of course section templates of the training course, and according to learning planning information contained in the target course section template and the learning behavior data, determining a target course section template corresponding to the target course section; and determining learning progress information of the target student user for the target course section. According to the technical scheme provided by the embodiment of the invention, the efficiency and accuracy of training course learning progress statistics can be improved.

A method of predicting a composition of a meal includes obtaining data relating to concentration of a biomarker of an individual that consumed the meal, analyzing the data to determine the composition of the meal, wherein the analyzing comprises using a computational model. The computational model relies upon statistical learning techniques to breakdown meal composition into macronutrient levels.

The invention belongs to the technical field of radar signal processing, and discloses a low-altitude surveillance radar target intelligent identification system. The method comprises the following steps: analyzing and standardizing radar original echoes or detection output, associating discrete plots based on space-time correlation to form a target track, and generating track motion state parameters; performing space-time-frequency combined feature extraction on the trace point and / or track sequence, and constructing a comprehensive feature set comprising one or more of rotor micro Doppler, multi-scale time sequence, time-frequency non-stationary and physical constraint; the features are input into a multi-expert dynamic fusion recognition model, the model at least comprises time sequence modeling experts and statistical learning experts, and output of all the experts is fused in a self-adaptive mode through a dynamic weight fusion mechanism; and performing track-level aggregation and stabilization processing on the prediction result, and outputting a track-level target category label and confidence, thereby avoiding misclassification caused by a single model under the condition that the track is incomplete, and enhancing the stability of an identification result in a complex environment.