21results about How to "Avoid underfitting" patented technology

The invention provides a stiffened wallboard life prediction method based on a data enhanced physical information neural network, which comprises the following steps: determining influence parameters of crack propagation fatigue life of a stiffened wallboard and a discourse domain of the influence parameters, obtaining sample points through sampling, calculating a stress intensity factor and fatigue life based on a finite element model, and constructing a training set; a double-layer physical information neural network is established, a first sub-model predicts a stress intensity factor by taking an influence parameter as input, a second sub-model combines the influence parameter and a stress intensity factor prediction value, a physical loss function based on a Paris crack propagation law is introduced, training is performed in combination with a data loss function, and fatigue life prediction is realized; and representing the influence parameters as fuzzy variables, obtaining a multi-level-cut set, and predicting the fuzzy response of the fatigue life by using the trained neural network. According to the method, a physical mechanism and data driving can be effectively fused, and efficient prediction of the crack propagation fatigue life of the stiffened wall plate under the uncertain condition is achieved.

The application provides a provincial meteorological and administrative collaborative wind speed prediction method based on dynamic federal learning, relates to the cross technical field of artificial intelligence and energy meteorology, S1, basic data of N meteorological monitoring stations in a province is acquired, is divided into M initial subgraphs according to the administrative boundary of a prefecture-level city and is allocated an edge server, each collaborative subgraph pair is determined according to the Pearson correlation coefficient of the wind speed sequence of the station, the time sequence data deviation of the synchronized subgraph is synchronized through a dynamic time warping algorithm after every 3 rounds of federal iteration, and the subgraph data after synchronization is output. Through an endogenous decision and optimization framework, three links of dynamic coordination of cross-domain collaboration, local learning and gradient transmission are coordinated, and finally a global optimal provincial meteorological and administrative collaborative wind speed prediction method based on dynamic federal learning is achieved among multiple targets such as prediction accuracy, convergence speed and communication cost.

The application relates to the computer technical field and discloses a modeling method of a statistical mixed model in a big data distributed scene. The method comprises the following steps: distributing and storing data shards; initializing model parameters; iteratively performing an expectation step and a maximization step; the expectation step is scheduled to a GPU node to perform parallel calculation on posterior probability; the maximization step is scheduled to a CPU node to aggregate statistics and update parameters, perform component merging / deletion, and perform convergence judgment; meanwhile, a memory reuse mechanism based on reference counting and scope analysis is adopted to reduce redundant data transmission. Through heterogeneous task scheduling and memory collaborative optimization, the training speed, resource utilization rate, and model self-adaptation capability are improved.

The application provides a ship type classification prediction method and system based on K-means and XG-Boost, acquires ship data and carries out pretreatment, then adopts a K-means clustering algorithm to respectively cluster each kind of data in the pretreated ship data to obtain multiple clusters, calculates error sum of squares of all data in each cluster, calculates the classification number of each kind of data according to the error sum of squares, selects a certain classification number by using an elbow method, marks the ship type of all clustered ships according to the classification number, then takes the ships with marked ship types as training set samples, adopts an XG-Boost classification algorithm to train the training set samples to obtain multiple classification prediction models, verifies the multiple classification prediction models to obtain an optimal classification prediction model, and predicts the ship types of all ships in the world according to the optimal classification prediction model and marks the ship types. The application can accurately classify all ships in the world, and can avoid overfitting and underfitting of the model while ensuring the accuracy.

The application relates to the technical field of microwave remote sensing, and proposes an adaptive wet troposphere path delay inversion method and system fusing wind speed overlapping partition modeling, multi-scale collaborative sample balancing and model soft fusion. The method comprises the following steps: acquiring multi-source data and preprocessing the multi-source data to obtain to-be-inverted samples; according to a preset threshold, the full-range wind speed of the to-be-inverted samples is divided into a plurality of wind speed subintervals with overlapping transition zones; the to-be-inverted samples are input into trained inversion submodels, the output results of the inversion submodels are weighted and summed according to final fusion weights, and finally, wet troposphere path delay inversion results are obtained; when abnormal wind speed or invalid submodel prediction output is detected, the output result of a pre-trained global unified model is taken as the final inversion result. While guaranteeing overall inversion precision and stability, the application can significantly suppress systematic deviation under sea conditions of extremely low and extremely high wind speed.

This invention discloses an intelligent detection method and system for bid format compliance of sealed bid documents, relating to the field of electronic bidding. The method includes: performing format detection based on the format rules of the bidding documents to obtain a set of violation records and a single format detection report; constructing a format error fingerprint vector, calculating cross-file error similarity, and generating a cross-file error similarity matrix; constructing an initial suspected association graph using each sealed bid document as a node and the similarity values ​​in the cross-file error similarity matrix as edge weights between nodes; dynamically adjusting the connectivity threshold, updating the node connection relationships in the weighted disjoint-set data structure, and defining the groups divided under the optimal connectivity threshold as abnormal similar file groups; calculating a bid-rigging suspicion score, generating a bid-rigging detection report, and outputting the bid compliance detection results. This solves the technical problem in existing technologies where isolated compliance judgments on individual bid documents fail to identify potential bid-rigging behavior.

The application discloses a multi-modal depth perception and grasping system based on transparent objects, and relates to the field of robot operation, comprising a multi-spectral perception module, a depth correction module, a grasping posture generation module and a control module. The application comprehensively acquires visual information and thermal radiation information of the transparent object by combining two perception methods of RGB-D image and thermal imaging (TIR) image, and analyzes systematic errors of the depth map to detect error sources of the RGB-D camera and the TIR camera. The system adopts an encoder-decoder model as a depth correction core, the model extracts complementary features of the RGB-D image and the TIR image through a modal exclusive encoder, and utilizes a feature fusion module to complete feature alignment and integration, thereby effectively improving the depth estimation accuracy on the transparent surface. Meanwhile, a Bayesian optimization method is adopted to optimize hyperparameters of the depth correction model, through hyperparameter optimization and model fitting processing, the system effectively avoids overfitting and underfitting problems, and further improves the robustness of the depth correction model and the grasping accuracy of the transparent object.

The application discloses a feature marker combination based on PET / CT molecular imaging and a human body aging clock evaluation method, and belongs to the technical field of medical data. The technical problem to be solved is to solve the defects of the existing human body aging evaluation method, such as insufficient comprehensiveness and limited accuracy. The technical solution points are to provide a feature marker combination composed of functional feature markers and volume feature markers; a support vector machine regression (SVR) algorithm is used to construct a model, the mean absolute error (MAE) in the independent verification set is 5.68 years, the determination coefficient R 2 is 0.85. The application can be used for human body aging clock evaluation, and provides a new perspective of aging evaluation based on multiple organ perspectives and verifiable marker support.

This application discloses a method, apparatus, equipment, and medium for assessing voltage sag fault risk, belonging to the field of voltage sag fault risk assessment. The method involves: collecting test points of the withstand characteristic curves of the electrical equipment under test under different voltage sag initiation angles and load conditions; each test point includes residual voltage and duration values; constructing multiple first Gaussian mixture models with different numbers of Gaussian components based on the test points; iteratively solving the optimization parameters of each model using the expectation-maximization algorithm until convergence, obtaining multiple second Gaussian mixture models; selecting a target Gaussian mixture model using the Bayesian information criterion; and finally, inputting the voltage sag test points to be assessed into the target Gaussian mixture model to obtain the assessment result characterizing the voltage sag fault risk of the electrical equipment under test. Therefore, by implementing this application, the problem of low accuracy in voltage sag fault risk assessment in the prior art can be solved.

This invention provides an anomaly early warning method for natural gas purification plants. The method includes: collecting historical and real-time data from various equipment within the purification plant based on the frequency of abnormal conditions and performing preprocessing; analyzing and selecting feature parameters from the preprocessed data and establishing a rule table; inputting the preprocessed real-time data into a data training model for classification and training to obtain classified abnormal operating condition data; establishing a multi-source data fusion algorithm model, inputting the classified abnormal operating condition data into the model for calculation to obtain the category and probability of the abnormal operating condition; and performing judgment and analysis on the obtained categories and probabilities of the abnormal operating conditions based on the rule table to achieve accurate monitoring and early warning of abnormal states of equipment within the purification plant. This invention introduces temporal analysis, using temporal convolutional networks and attention mechanisms to improve recognition accuracy; simultaneously, it proposes an imbalanced classification algorithm, improving the accuracy and efficiency of anomaly early warning for natural gas purification plants.

This invention provides a method for converting resistivity logging to sonic velocity logging. Based on a defined rock physics model, it utilizes the relationships and patterns between different physical parameters to establish an analytical expression between rock porosity, resistivity, and velocity, obtaining a discrete mapping relationship between resistivity and P-wave velocity. By continuously optimizing the rock physics parameters, it obtains the mapping curve between resistivity and P-wave velocity, avoiding underfitting caused by cross-sectional analysis and linear fitting of rock physics parameters alone. This reliably achieves an effective conversion from resistivity logging to sonic velocity logging. This invention can support pre-stack depth migration velocity updates based on logging information from drilled sections, predict seismic velocities ahead of the drill bit, and enable more accurate seismic imaging and drilling decisions.

The invention discloses a neural network model training method and an ellipsometry method, and relates to the technical field of ellipsometry, the method comprises the following steps: obtaining a reference spectrum and a to-be-calibrated spectrum corresponding to the reference spectrum, the reference spectrum being obtained by measuring a sample through a reference ellipsometer, and the to-be-calibrated spectrum being obtained by measuring the sample through a to-be-calibrated ellipsometer; the to-be-calibrated spectrum and the simulation spectrum serve as training input, the to-be-calibrated spectrum and the simulation spectrum serve as training labels, a neural network model is trained, a loss function of the neural network model is converged, the trained neural network model is obtained, the loss function comprises a first loss item and a second loss item, and the first loss item and the second loss item correspond to each other; the first loss item is obtained based on the difference between the training output of the spectrum to be calibrated and the corresponding reference spectrum, and the second loss item is obtained based on the difference between the training output of the simulation spectrum and the simulation spectrum. The model can adjust the measurement spectrum of the ellipsometer to be calibrated so as to eliminate the measurement difference of the ellipsometers to be calibrated.

The application discloses a model training method and device based on multimedia resources, equipment and a storage medium. The method comprises the following steps: determining a first characteristic vector of a plurality of training samples through a first model, and then performing model optimization on the first model based on the first characteristic vector of the plurality of training samples; determining an object characteristic vector of a target training sample and a resource characteristic vector of a candidate multimedia resource in a database through the optimized first model; distilling one or more candidate multimedia resources from the database based on the matching degree between the resource characteristic vector of each candidate multimedia resource and the object characteristic vector of the target training sample; constructing one or more distilled samples by using the object characteristic information of the target training sample and the distilled candidate multimedia resources respectively; and performing model training on a second model by using each distilled sample and the plurality of training samples. The application can solve the underfitting problem of the model in a small data volume scene and improve the generalization ability of the model.

The invention belongs to the technical field of deep reinforcement learning, and discloses a multi-agent multi-task collaborative reinforcement learning method based on a space-time fusion architecture, and the method comprises the steps: 1, initializing a task sampling probability, and forming an entity embedding vector sequence and a task embedding vector; step 2, inputting the entity embedding vector sequence and the task embedding vector into an anti-noise feature extraction layer to obtain deep time sequence features; 3, generating a dynamic weight matrix in real time, and obtaining a local Q value output by the intelligent agent; 4, the Transform hybrid network receives a global state vector, a task embedding vector and a local Q sequence of the environment, and gives a global action value; 5, calculating the total loss, and 6, based on the total loss, training an optimization unit to update whole network parameters, and meanwhile, judging whether a preset evaluation round is reached or not. The method has higher anti-interference capability and more stable control performance, and realizes adaptive control of multiple heterogeneous tasks.

The application relates to the technical field of water level prediction, and particularly discloses a basin water level prediction system and method for complex terrains, wherein the method is characterized in that: a feature screening algorithm is used to screen a preliminary model feature set, and effective features which have a significant influence on the transit of a flood peak and the risk of backflow are reserved; the effective features are scientifically divided into a training set and a test set; the training set is used to enable the model to fully learn the mapping relationship between the buffer state of a buffer confluence unit, the flood discharge, the movement of the flood peak and the change of the confluence water level under complex terrains; and the fusion of hydrological data and topographic image data covers key influence factors in the time and space dimensions on the data level; the water flow lagging effect is captured through an upstream water level lagging feature set on the feature level; the discharge position is accurately positioned in combination with the discharge channel and the buffer state of the buffer confluence unit; and the transit path of the flood peak and the peak water level are simulated, so that the accurate simulation of the transit of the flood peak in the complex terrain and the reliable prediction of the backflow risk can be realized.

The application discloses a kind of semi-supervised RGB-D object classification methods based on angle prediction pre-training, comprising: the rotation angle predictor of RGB and depth image is trained, obtains the network model by unsupervised training;The feature extraction part of rotation angle predictor is used as feature extractor;Object class predictor of RGB and depth image is constructed;Object class predictor is trained using image with label, and the result of RGB image or depth image semi-supervised classification is obtained;Then the prediction result of object class predictor is fused using the complementary information in RGB and depth image, the parameter of feature extractor is fine-tuned, so that the feature extraction part of rotation angle predictor adapts to the object classification task based on RGB image, depth image or RGB-D image.The application improves performance by depth mutual learning and fuses two kinds of modal specific object class predictor, after mutual learning, the object classification accuracy of RGB and depth image has been significantly improved.

The present application relates to the technical field of point cloud data processing, and more particularly to a self-adaptive grid fitting point cloud filtering method and system based on gradient compensation. The method divides the grid by the local density of the point cloud and performs surface fitting, calculates the grid gradient variation to determine the gradient compensation, and then corrects the fitting surface parameters. Based on the corrected parameters, the deviation of each point from the fitting surface and the local curvature are calculated to determine the adaptive filtering threshold, thereby classifying and removing the noise points. The present application can effectively maintain the detailed features of the point cloud and improve the filtering robustness of complex surfaces and noise.

This application provides a method, apparatus, and device for online real-time source apportionment of VOCs. The method includes: screening atmospheric VOCs receptor samples to obtain a first receptor data matrix; when it is determined that there is no local source component spectrum corresponding to the receptor sample, performing a source-free spectrum analysis process including: performing PCA on the first receptor data matrix, obtaining principal components with eigenvalues ​​greater than 1 in each source component, calculating the cumulative contribution rate of eigenvalues ​​of N principal components, determining the number of first sources as the number of factors based on the cumulative contribution rate of eigenvalues, running PMF to decompose the first receptor data matrix to obtain a source component spectrum matrix; calculating the correlation coefficient and mass percentage based on the source component spectrum matrix to obtain the number of second sources; identifying sources that meet the number of second sources based on a preset pollution source characteristic tracer material library and source component spectrum matrix, obtaining source identification results, and performing trend analysis using the source identification results.

The invention discloses a water turbine thrust pad temperature prediction method based on working condition self-adaption, and relates to the technical field of hydroelectric generating set operation and maintenance monitoring and intelligent diagnose.The method comprises the steps that a working condition boundary fitting model is built based on the corresponding relation between a water head and active power according to historical operation interval data of a set; slicing an original sequence by adopting a rolling window mode, and then automatically extracting derivative features from each window by utilizing a feature derivation tool; constructing training data sets, grouping the training data sets according to different working conditions, and constructing an XGBoost regression model; selecting a corresponding XGBoost prediction model according to different working conditions, and performing anti-standardization processing on the output predicted tile temperature to obtain an actual temperature predicted value; and calculating a residual error between the predicted value and the real thrust pad temperature, and carrying out statistical analysis on the residual error in a set time window. According to the method, the XGBoost regression model is independently trained, so that the prediction robustness, the model training efficiency and the prediction stability under a complex operation condition are improved.

The present application relates to the technical field of gene editing and bioinformatics, and particularly to a DNA repair prediction method based on sample difference adaptation, comprising the following steps: obtaining sample data containing a target sequence, a cleavage site and context information thereof, and DNA repair output statistical data corresponding to the sample; processing the output statistical data to obtain label data for training; then constructing input representation based on the target sequence, the cleavage site and the context information thereof, and establishing a DNA repair output prediction model to predict the DNA repair output probability distribution of a sample to be tested. The present application proposes a DNA repair prediction method based on sample difference adaptation, the core starting point of which is to introduce a learnable adaptive adjustment mechanism under the premise of maintaining mechanism interpretability, so that the model can strike a balance between mechanism constraints and distribution adaptation capability.

The application provides a feedback and estimation method for dynamic response surface of a space tensioned membrane structure, relates to the technical field of dynamic response and state estimation, and comprises the following steps: collecting tensioning force of a rope of the membrane structure, inclination of the rope relative to a balance plane, swing angular velocity of the rope and frame strain through a boundary sensor array, filtering and normalizing the collected signals, and constructing a high-dimensional boundary dynamic observation vector; the application adopts an adaptive surface estimation mechanism without prior model driving, learns dynamic characteristics of the structure through real-time boundary data online, and corrects parameters, so that it is unnecessary to previously establish a finite element model or extract a fixed modal shape, thereby effectively avoiding model mismatch problems caused by factors such as temperature alternation, pre-tension relaxation and structural micro-damage in orbit, significantly improving long-term robustness of surface estimation in orbit, and meeting stable working requirements of the space large tensioned membrane structure under variable working conditions.