62results about How to "Learn accurately" patented technology

The invention provides an automatic task allocation method and system for an elevator emergency rescue platform, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining elevator operation state data and environment information data, constructing an emergency task model, and obtaining classified emergency tasks; inputting the classified emergency tasks into a deep reinforcement learning model to generate a task allocation scheme; the deep reinforcement learning model is formed by combining a deep Q network structure with a priority scanning algorithm, and a priority experience playback mechanism is introduced; and constructing a multi-objective optimization module comprising a time constraint, a space constraint and a resource constraint, and optimizing the task allocation scheme by the multi-objective optimization module based on an improved particle swarm algorithm to obtain an optimal task allocation scheme.

The invention discloses a mobile phone screen glass defect detection method based on improved YOLOv8, and relates to the technical field of mobile phone screen glass cover plate quality detection. The method comprises the following steps: simulating an industrial production environment, collecting mobile phone screen glass images, and constructing a data set; improvement and optimization are carried out on the basis of the YOLOv8 target detection network model, a mobile phone screen glass defect detection classification model is constructed, and parameters are trained and optimized; constructing a semantic segmentation model of the YOLOv8 network, segmenting the defects of the mobile phone screen glass image to be identified, and realizing detection and identification of different defect category edges; and fitting the actual damage size of the mobile phone screen glass defect range. The mobile phone screen glass defect detection system is suitable for accurate detection of mobile phone screen glass defect characteristics, can measure the defect range, and provides intelligent and efficient services for the mobile phone screen glass quality inspection process.

The invention belongs to the technical field of federated learning, and more specifically relates to a heterogeneous federated learning-oriented adaptive differential privacy protection method. The method includes introducing adaptive noise at different levels of the model. The contribution of each level of the model to the overall learning effect is different, and in order to reduce damage to key features as much as possible while noise addition is performed, differential noise addition is performed on different parts based on the importance degree of the levels of the model, that is, less noise is applied to the more important level, and more noise is applied to the secondary level. According to the method, the problem that the traditional differential privacy federal learning method usually has negative influence on the performance of the model due to the introduction of noise, especially the problem that the convergence speed and precision of the model are reduced is solved.

The invention provides an optimization method based on an industrial energy consumption big data model. The optimization method comprises the following steps: S1, acquiring a reference data set; s2, a target function with the purpose of minimizing the total energy consumption within the unit industrial production time is determined, and construction of an energy consumption prediction model embedding layer is completed; s3, enriching data background and influence factors through a real-time energy consumption data set Oj collected from a monitoring station j set in the current industrial production cycle, and realizing real-time calibration and display of the fine-tuned model by establishing a contribution matrix; and S4, forming an energy consumption point set. According to the method, the bag-of-words model is constructed by formulating the constraint conditions, so that the construction and output of the bag-of-words model can meet the constraint requirements of the optimization scheduling model, the obtained model input data is text data with constraint, and the subsequent energy consumption model training and prediction process not only considers the statistical characteristics of energy consumption data, but also considers the energy consumption data. And the computing power constraint in the prior art is reduced.

The invention discloses a lightweight crack penetration degree detection method based on ground penetrating radar and Swin Transform optimization, and the method comprises the following steps: 1, collecting ground penetrating radar data, and carrying out the preprocessing and feature extraction; step 2, constructing a multi-feature fusion data set; step 3, designing a two-stage through crack detection model based on Swin Transform optimization; 4, a first-stage through crack detection model optimized on the basis of Swin Transformer-YOLOv8 is designed, and a second-stage through crack detection model is designed on the basis of Swin Transformer-YOLOv8 optimization; and step 5, training and predicting the data of the training set and the test set according to the model architecture in the step 3 and the step 4. According to the invention, the accuracy of crack penetration degree detection is improved, and a scientific basis is provided for road maintenance; the detection speed is greatly improved, and the requirement of large-area rapid detection is met; the calculation complexity is reduced, the equipment performance pressure is reduced, and the detection efficiency is improved; the generalization ability of the model is enhanced, and the method can adapt to different road conditions; and the development of an intelligent road detection technology is promoted.

The invention provides a multi-light-source scattering water turbidity detection method, system, equipment and medium, and relates to the technical field of turbidity detection.The method comprises the steps that the influence of wavelength combinations on pollutant recognition performance is evaluated according to a preset fitness function, and an optimal wavelength combination is obtained; aiming at the optimal wavelength combination, adopting a reward function to learn a light source switching and irradiation strategy, and determining a light source control strategy; controlling a multi-wavelength light source to irradiate the water sample by adopting the light source control strategy to obtain scattered spectrum data; performing wavelet multi-scale decomposition and soft threshold filtering on the scattering spectrum data to obtain preprocessed scattering spectrum data; and establishing a quantitative relation model by adopting the preprocessed scattering spectrum data, and determining the pollutant concentration. Through wavelength combination optimization and intelligent learning of a light source control strategy, the accuracy and efficiency of water quality pollutant detection are remarkably improved, and an intelligent solution is provided for water quality monitoring.

The invention relates to the technical field of machine learning and software development, in particular to a software development system based on a machine learning algorithm, which comprises an acquisition processing unit, a model training unit, an interaction prediction unit and an optimization feedback unit, and is characterized in that the acquisition processing unit constructs a database, extracts key elements such as variable names and converts the key elements into vectors; the model training unit adopts recurrent neural network hierarchical training and gradually introduces intermediate and complex structure codes and learns logic relation and semantic relation from code snippets related to a basic grammar structure, the interaction prediction unit screens and extracts the code snippets according to code function module labels and historical use frequency, the code snippets are input into a model after vector conversion, and the model is used for prediction. A probability weighting algorithm is used for predicting subsequent fragments, an optimization feedback unit calculates the similarity between predicted fragments and actual fragments, an optimization mechanism is automatically triggered if a threshold value is not reached, parameters are adjusted by utilizing reinforcement learning, and prediction accuracy is used as a key part of a reward function so as to improve the system performance.

The invention discloses an infectious disease prediction and transmission link speculation method and system based on a random forest, and belongs to the technical field of infectious disease transmission risk prediction, and the specific steps are as follows: data acquisition and preprocessing: collecting case information from official websites and social media platforms of a plurality of provincial and urban health departments throughout the country, cleaning and preprocessing the data set; predicting the illness time of the patient: constructing and training a model by using a random forest algorithm, and accurately predicting the infection time of the patient; evaluating an algorithm prediction result, including evaluating model performance through classification prediction accuracy and correction accuracy; the disease transmission relation speculation comprises the steps of constructing a dynamic transmission model based on the predicted infection time and generating an infection time sequence diagram, the transmission path of diseases in urban population is revealed, and a scientific basis is provided for infectious disease control and public health decision making.

The invention provides a clean coal ash content control method and system in a dense medium separation process. The method comprises the following steps: collecting historical data in the dense medium separation process; preprocessing the data; constructing and training a state estimation model and a system dynamic model; and performing real-time control on the ash content of the clean coal based on the constructed state estimation model and the system dynamic model. According to the method, the modeling problem of the large time-lag characteristic in the dense medium separation process is effectively solved, the influence of the feeding pressure on the clean coal ash content is comprehensively considered, the model can more accurately learn the bottom layer physical characteristics in the dense medium separation process, proper control actions can be adopted in time to compensate the influence of external interference factors, the response speed is high, and the real-time performance is high. The method has strong anti-interference capability and is insensitive to noise of ash content measurement. When the scheme is applied to production, the quality of a coal product can be improved by accurately controlling the ash content of the clean coal; and automatic control can be fully achieved, the labor intensity of operators is relieved, and the production efficiency is improved.

The invention belongs to the technical field of histopathology full-slice image analysis, and particularly relates to a histopathology full-slice image classification method and system based on a position coding converter, and the method comprises the steps: obtaining a pathological full-slice data set with an image and a class label thereof, and carrying out the segmentation preprocessing; using a feature encoder to extract block image features, and establishing a converter multi-instance learning network model based on position coding guidance; training a network model by using the pathological full-slice data set containing the weak label; according to the method, the problem that detailed annotations of pathological images are difficult to obtain can be solved through a multi-instance learning weak supervision method, and the importance of spatial and semantic relationships among different sub-regions in the pathological sections can be effectively considered; and completing a classification task of pathological full slices and carrying out visual interpretive analysis.

The invention discloses an image enhancement method and device based on double-teacher knowledge distillation and a storage medium, and belongs to the technical field of digital image processing. According to the method, an underwater image enhancement network for double-teacher knowledge distillation is provided, and an image processing problem is divided into a plurality of small problems by the framework; a Freq-UIE network and a vision-based Fourier processing module are designed, and frequency domain information is added to different layers of the network, so that the effect of edge and detail enhancement is achieved; in order to better match and distill features of multiple teachers and students, a knowledge collection and distillation module is provided, and the features of the students and the multiple teachers are processed on each scale. Experimental results prove that compared with the prior art, the method has the advantages that the detail and texture information of the image can be well enhanced, the color recovery is more accurate, and the edge is clearer.

The invention discloses a mine water end water quality regulation and control system which comprises a data acquisition module used for acquiring water quality parameters of mine water; the model construction module is connected with the data acquisition module and is used for constructing a water quality prediction model based on the recurrent neural network and the historical water quality data; the water quality prediction module is connected with the model construction module and is used for analyzing the water quality parameters through a water quality prediction model to obtain a water quality change trend; and the regulation and control module is connected with the water quality prediction module and used for regulating and controlling water quality based on the water quality change trend. According to the invention, data acquisition, model prediction and automatic regulation and control are integrated, automation and intellectualization of water quality regulation and control at the tail end of mine water are realized, and the need of manual intervention is reduced.

The invention provides a speech synthesis model training method, a speech synthesis method, a speech synthesis device and speech synthesis equipment, and belongs to the technical field of computers. The first speech synthesis model comprises a language network and a diffusion network; the method comprises: processing a sample text and a sample audio through a language network to obtain a first audio feature, the sample text and the sample audio being consistent in content, the first audio feature comprising a content feature of the sample text and an acoustic feature of the sample audio; performing noise prediction on the sample audio added with the reference noise based on the first audio feature through a diffusion network to obtain predicted noise; determining a first training loss based on the reference noise and the predicted noise; and updating parameters of a language network and a diffusion network in the first speech synthesis model based on the first training loss to obtain a second speech synthesis model. According to the technical scheme, the purpose of end-to-end training of the speech synthesis model is achieved, the training steps are simplified, and the training efficiency of the model is improved.

This application discloses a data learning method, data learning device, and medium for elevator shafts, applicable to the field of elevator technology. The method includes: based on the obtained information about the length of the magnetic shielding plate within the elevator shaft, stopping the elevator at a preset position in the leveling zone; controlling the elevator to move to another floor corresponding to the target floor; during the elevator's operation, detecting the first signal conversion of the elevator's leveling switch signal to determine the departure position from the target floor; controlling the elevator to continue moving from the departure position; based on the signal conversion of the elevator's up / down forced switch signal, determining the trigger position of the elevator's up / down forced switch; detecting the second signal conversion of the elevator's leveling switch signal to determine the arrival position at the other floor; and determining the floor distance between the two floors based on the departure position of the target floor and the arrival position of the other floor. In the case of two floors, there is no need to shorten the length of the magnetic shielding plate, accurately learning the elevator shaft data.

The invention relates to the field of artificial intelligence and financial science and technology, and discloses a bank data resource checking method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring data resources and preprocessing the data resources to obtain sample data; building a training sample according to the sample data; performing category prediction on the training sample based on a large model to obtain a prediction result; evaluating the prediction result, and outputting a final data item according to the prediction result; and mapping the final data item to a data item corresponding to a data resource directory. According to the method, a large model technology is adopted to perform category prediction on the training samples, so that the limitation that traditional machine learning and deep learning have insufficient understanding on text information can be overcome, more accurate learning and prediction on the training samples are realized, the accuracy of a data resource inventory result is effectively improved, and the accuracy of the data resource inventory result is improved. Bank data resource checking is promoted to develop towards a more intelligent and automatic direction.

The application discloses a bucket tooth health assessment method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring state data of multiple bucket teeth on a bucket; mapping the multiple bucket teeth as nodes in a graph structure respectively, constructing a physical adjacent edge according to the spatial adjacency relationship among the multiple bucket teeth, and constructing a symmetric coupling edge according to the symmetry relationship among the multiple bucket teeth about the center of the bucket, so as to establish a bucket tooth group compound graph structure containing the physical adjacent edge and the symmetric coupling edge; initializing the node features of the corresponding nodes in the bucket tooth group compound graph structure according to the state data of the multiple bucket teeth, and initializing the edge features of the edges in the bucket tooth group compound graph structure according to the spatial geometric relationship and the load transmission relationship among the multiple bucket teeth; inputting the bucket tooth group compound graph structure, the node features of each node and the edge features of each edge into a graph neural network model, and outputting the failure probability of the multiple bucket teeth and the overall health degree of the bucket. The application improves the accuracy and comprehensiveness of the bucket tooth health assessment.

The invention relates to the technical field of reinforcement learning, in particular to a bunching station parameter optimization method based on extended Kalman filtering and reinforcement learning, and the method comprises the steps: randomly selecting an initial state from an environment based on an initial strategy and a preset target state, carrying out the sampling, and collecting a plurality of pieces of trajectory data composed of continuous empirical tetrads; training a probabilistic neural network by using the collected trajectory data in the first round of sampling to obtain a state transition model; carrying out extended Kalman filtering on each piece of trajectory data in combination with the state transition model, replacing the filtered state at the next moment into an experience tetrad of each piece of trajectory data, and storing the filtered state into an experience playback pool; and randomly sampling the experience tetrad from the experience playback pool by using a DDPG algorithm, learning and updating the strategy to obtain a new strategy, and repeating the steps until the strategy learning is completed. According to the method and the device, the influence caused by system errors can be relieved, the accuracy of state estimation is improved, and thus the strategy learning is more accurate.

The invention provides an image noise reduction model training method based on a UNet network, a computer device and a storage medium. The method comprises the following steps: acquiring original image data in different scenes; generating a noise-true value image pair according to the original image data; establishing a teacher model and a student model by using a UNet network; training a teacher model and a student model by using the noise-truth value image pair, and outputting a final image noise reduction model; wherein the step of training the teacher model and the student model by using the noise-truth value image pair comprises the following steps of: training the noise-truth value image pairs corresponding to a preset number of gain multiples by using the teacher model to obtain a sub-teacher model under each gain multiple, and performing knowledge distillation on the student model by using the preset number of sub-teacher models to obtain a final image noise reduction model. According to the invention, the image noise reduction model has a good noise reduction effect and is convenient to deploy on low-computing-power equipment.

The present invention provides an optimization method based on an industrial energy consumption big data model, comprising S1, obtaining a benchmark data set; S2, establishing a target function with the goal of minimizing the total energy consumption per unit industrial production time, and completing the construction of an energy consumption prediction model embedding layer; S3, collecting a real-time energy consumption data set O from a monitoring site j set in the current industrial production cycle. j Enrich the data background and influencing factors, and achieve real-time calibration and display of the fine-tuned model by establishing a contribution matrix; S4, forming a set of energy consumption points. The present invention constructs a bag-of-words model by formulating constraint conditions, which can ensure that the construction and output of the bag-of-words model can meet the constraint requirements of the optimization scheduling model. The obtained model input data is text data with constraints, which ensures that the subsequent energy consumption model training and prediction process not only considers the statistical characteristics of the energy consumption data, but also reduces the computing power constraints in the prior art.

The present invention discloses a traffic flow prediction method based on spatial partitioning and static weighting, which can quickly partition the data space corresponding to the historical observation data of traffic flow and realize the association mapping of observation records to data subspaces, and can measure the specificity and outliers of observation records according to the distribution characteristics of observation records in data subspaces and assign differentiated value weights, and realize accurate prediction of traffic flow in combination with long short-term memory neural network models. Since this method uses static weighting technology, it can complete the calculation and storage of the value weights of observation record data at one time in the preprocessing stage, with low calculation cost, and avoid additional time overhead for the training process of long short-term memory neural network models. The use of long short-term memory neural network models combined with spatial partitioning and static weighted data processing technology reduces the impact of invalid or abnormal observation data on model performance, and effectively improves the accuracy of traffic flow prediction.

The method belongs to the technical field of panoramic monitoring of ultra-high voltage converter station, and aims to solve the problem that panoramic monitoring image data is directly uploaded to the cloud, occupying a large amount of cloud resources. By adopting multi-scale convolution blocks in the deep multi-scale residual network model to construct low-order and high-order features of images of various scales, the incomplete phenomenon of image detail extraction is avoided, and the residual learning mechanism is adopted to retain low-order rough features, thereby improving the reconstruction ability of the image. Topology optimization constructs the topology structure of the heterogeneous network, and the framework combining deep reinforcement learning and Monte Carlo tree search is used to construct the network according to the pre-defined topology rules. The search result of the Monte Carlo tree strengthens the learning of the deep convolutional neural network, so as to obtain more accurate prediction in the next iteration. After the data is processed in the edge side, it is transmitted to the cloud storage, saving the cloud storage space and transmission bandwidth.

The invention discloses a data set generation method and system for converter harmonic coupling feature learning, and the method comprises the steps: determining the access capacity and position of a power electronic converter in a power distribution network, randomly generating m groups of multi-frequency harmonic voltages under the amplitude limitation, and defining the distance between two groups of harmonic voltages; calculating to obtain a distance matrix among all harmonic voltage data, and according to a traversal rule, carrying out small signal superposition on m groups of data in sequence; obtaining a steady-state equation coefficient matrix; dividing the harmonic voltage at the converter as large disturbance into a plurality of stages, and calculating the small disturbance response of each state variable of the small signal equation under the harmonic voltage disturbance at each stage; superposing the small disturbance response of each state variable solved in the stage into a steady-state component of the state variable, and updating a small signal equation set of the system; and superposing the disturbance of the next stage, repeating until the amplitude of the harmonic voltage disturbance is the same as the set target, and finally obtaining the harmonic coupling characteristics of the converter under m groups of random harmonic voltage data.

The invention relates to a language learning system, which comprises a learning system, a video player and a foreign language video material, and is characterized in that the output end of the learning system is electrically connected with the video player, and the foreign language video material is in data connection with the learning system and the video player through a data memory. By adopting the translation algorithm module and the subtitle display module, an English learning program is highly simplified, English learning becomes closer to a native language learning process, information of a foreign language video material is decomposed, extracted, translated and combined, a learner can comprehensively and accurately learn the foreign language video material, and meanwhile, the learning efficiency is improved. The translated subtitle information is transmitted to the video player in real time to be displayed, so that a learner can view the translated subtitle information while watching the video, and the learning efficiency and effect are improved.

The invention relates to a discrete probability prediction method and device based on a conditional variation auto-encoder, and the method comprises the steps: constructing a discrete probability power prediction model based on the conditional variation auto-encoder and a decoder; obtaining historical power data and historical candidate feature data of target new energy power generation to establish a new energy power generation training data set; training a discrete probability power prediction model based on the new energy power generation training data set to generate a new energy power generation prediction model, and predicting target new energy power generation distribution by using the new energy power generation prediction model; and inferring actual Bernoulli distribution corresponding to the preset Bernoulli distribution according to the enhanced inference strategy and the trained decoder. Therefore, the problems that in the prior art, the prediction time cost of new energy power generation is high, meanwhile, the probability distribution of new energy is difficult to accurately and efficiently predict, complex mixed distribution of photovoltaic power generation is ignored, and balance between continuous and discrete photovoltaic distribution is difficult to obtain are solved.

The invention discloses a method and system for classifying cell division states based on latent space features. The method comprises the following steps: constructing a classification prediction network, comprising an encoder, an inverse division image generator, an abnormal division image generator and a normal division image generator; respectively collecting a plurality of images of inverse division, abnormal division and normal division as training sets for the three generators; fixing the encoder, respectively inputting a plurality of random vectors into the three image generators to generate a plurality of virtual images, obtaining virtual image sets of the three generators, and respectively training each generator using the training set and the corresponding virtual image set; fixing the generator, extracting latent space features from the training set using the encoder, respectively inputting the extracted features into the corresponding generators to obtain a generated image set, and training the encoder using the training set and the corresponding generated image set; and inputting a plurality of images of cells to be detected into the classification prediction network to obtain classification results of the division states of the cells to be detected.

The present invention provides a multi-head weakly supervised target detection method based on a cross-attention mechanism, comprising: acquiring an image, processing the image, and dividing the processed image into a training set and a test set; defining a class prototype feature, and constructing a weakly supervised target detection network WCAN model based on a cross-attention mechanism containing the class prototype feature; using a training data set to train the WCAN model; and based on the trained WCAN model, performing target detection on the test set image. Under the condition of only having a class label, the method can more comprehensively and fully perceive the position and class information, and realize more accurate and efficient target detection.

The invention relates to an intelligent leakage detection method based on a water supply pipeline sound signal, and relates to the technical field of water supply pipe network leakage monitoring. The method comprises the following steps: acquiring a sound signal of a water supply pipeline, and extracting an acoustic feature vector of the sound signal; the sound signals are manually labeled, a sound sample set with labels is obtained, and the labels comprise a first label representing leakage and a second label representing normal; and constructing a leakage probability model, and training the leakage probability model through the sound sample set. And inputting pipeline sound signals collected in real time into the leakage probability model, and judging whether the water supply pipeline leaks or is normal according to the predicted leakage probability output by the leakage probability model. The method has the effect of improving the recognition efficiency and accuracy of water supply pipeline leakage state detection.

The present invention discloses an electronic writing practice method based on a convolutional neural network, which belongs to the field of calligraphy. The electronic writing practice method based on a convolutional neural network is to learn the real writing effect based on the pen movement data of the result obtained by traditional electronic writing, establish the connection between the real pen and the capacitive pen, and apply the pen movement data to the traditional electronic writing method, so that the touch device can reflect more authentic writing content, thereby solving the problem that the writing of the capacitive pen cannot provide the effect brought by the writing of the real pen. The method can help the practitioner better understand his or her own writing practice level, master the pen movement rules, improve the writing method, and thus improve the calligraphy level.