99results about How to "Effective training" patented technology

The invention discloses a power system false data injection attack detection method and system. Measurement data of a power distribution system are acquired; constructing an attack sequence, and injecting the constructed attack sequence into the obtained measurement data to simulate a false data attack; constructing a false data injection attack detection model based on the expansion causal convolution and the recurrent neural network to analyze the state change of the power distribution system; and judging an attack behavior according to the analyzed state of the power distribution system. According to the method, time sequence data can be processed, the long-term dependency relationship and the complex mode can be captured, and the response capability to the complex attack mode can be improved, so that the false data injection attack can be effectively detected and prevented, and the safety and reliability of the power distribution system can be maintained.

The invention discloses a federal learning data processing method based on privacy protection, and the method comprises the following steps: S1, a client uses local data to train a generator of a generative adversarial network, and generates virtual data similar to the local data in distribution; s2, setting three privacy protection levels according to the privacy sensitivity of the local data; s3, performing model updating operation by using a homomorphic encryption technology; s4, utilizing a dynamic weight optimization mechanism based on a multi-arm bandit algorithm to obtain a priority processing data set; s5, preliminarily updating the global model, analyzing generated data through a reverse inference detection mechanism, and judging whether a privacy disclosure risk exists or not; s6, the server feeds back the updated global model to the client, and the client continues personalized model training in combination with local data; and S7, generating new virtual data, and carrying out the next round of global model training. According to the method, the generative adversarial network, differential privacy and homomorphic encryption are combined, so that efficient federated learning under privacy protection is realized.

The invention discloses pre-hospital emergency medical aid emergency help integrated equipment and system. The method comprises the following steps: S1, acquiring key vital sign data of a patient by using medical equipment; s2, calling basic information and medical history of the patient through equipment; s3, transmitting the collected data to a server of a remote data processing center in real time; s4, performing comprehensive analysis on the received vital sign data and the medical history information of the patient by applying a pre-trained deep learning model on the server, and generating a prediction report; s5, displaying a prediction result on a portable device or a remote monitoring station in real time by using a data visualization technology; s6, based on a visual prediction result and medical history analysis, guiding first-aid personnel to take targeted emergency medical measures; and S7, continuously monitoring the vital sign response of the patient and any health change related to the medical history, and updating the treatment strategy according to the further analysis of the deep learning model.

The invention belongs to the technical field of mushroom culture, and discloses a mushroom culture shelf lifting mechanism for a mushroom house, the mushroom culture shelf lifting mechanism comprises a rotating mechanism, the rotating mechanism comprises a first gear and a screw rod, bevel gears are fixedly connected to the sides, close to each other, of the first gear and the screw rod, and the screw rod is in threaded sleeve connection with a threaded sleeve; through cooperation of structures such as an inclined plate and a sliding rod, a rotating disc can conveniently rotate on a placement plate, after a first gear is meshed with a tooth groove, the rotating disc can drive a culture pot to rotate by 180 degrees and then continue to ascend and descend through transmission, it is ensured that each mushroom can receive light evenly, and the mushroom quality is improved. The illumination receiving time of different surfaces of mushrooms is kept consistent, so that the uniform growth of hyphae is promoted, the problem of excessive growth or uneven growth is avoided, meanwhile, different parts of a culture medium can obtain better conditions in the aspects of light receiving, atmospheric circulation and the like through rotation, uniform distribution and absorption of nutrients are promoted, and the growth rate of the hyphae is increased. The overall nutrition efficiency of mushroom growth is improved.

The invention provides a defect classification method and system for a semiconductor chip and a readable storage medium. The training method comprises the following steps: acquiring training set data, wherein the training set data comprises different types of defect images of the semiconductor chip; inputting the training set data into an initial defect classification model based on a deep convolutional neural network so as to train the initial defect classification model; in the training process, an attention mechanism module is introduced to the tail of at least part of convolution blocks of the initial defect classification model, and residual connection is introduced to at least part of the convolution blocks; and a target defect classification model is obtained after multiple rounds of iterative training. The method effectively balances model performance and computing resource requirements, and ensures that higher efficiency is kept while the classification precision is improved.

The invention belongs to the technical field of natural language processing, and discloses a text data labeling method and system based on weak supervised learning. The method comprises the following steps: performing pseudo label generation on a plurality of historical text data by using a weak supervised learning algorithm to obtain a plurality of source domain data provided with real labels and a plurality of target domain data provided with pseudo labels; constructing a text data labeling model by using a deep learning and adversarial training algorithm, and adjusting the text data labeling model by using a continuous learning algorithm; and labeling the real-time text data by using the adjusted text data labeling model to obtain labeled real-time text data, and updating the adjusted text data labeling model by using a continuous learning algorithm. The problems that in the prior art, manual labeling cost is high, labeling quality is uneven, and dependence on label data is high are solved.

The invention discloses a 3D target detection method and system based on binocular vision and open vocabularies, and relates to the technical field of target detection. The method comprises the following steps: calculating pixel features of each image by adopting an open vocabulary 2D detection model; pixel features of the left view and the right view are aggregated to each three-dimensional point; obtaining a parallax estimation map based on the pixel features of the left and right views; introducing a time aggregation guide parallax consistency constraint to obtain a depth parallax estimation map; converting the depth parallax estimation map into a 3D point cloud, and extracting features to obtain 3D features; and carrying out correlation integration on the text features, the 2D fusion feature vectors and the 3D features by utilizing object-level image guide detection loss, object-level cross comparison alignment and object-level self-comparison learning to obtain a three-dimensional target detection model, realizing three-dimensional target detection and distinguishing different target objects. According to the invention, accurate detection and distinguishing of the target can be realized, so that the flexibility and universality of 3D target identification are improved.

The invention relates to the technical field of edible mushroom planting, and particularly provides an edible mushroom planting moisturizing system. The edible mushroom planting moisturizing system comprises a planting cabinet assembly, a spraying assembly, a fan assembly, a linkage adjusting assembly, a temperature control assembly and a plurality of planting trays. According to the edible mushroom cultivation device, uniform humidity can be provided for edible mushroom cultivation, local over-wet or over-dry conditions are avoided, healthy growth of strains is facilitated, adjusting airflow is provided, uniform distribution of moisture among the planting trays is facilitated, meanwhile, redundant moisture can be taken away, mold breeding is prevented, and when the humidity is insufficient or too high during edible mushroom cultivation, the edible mushroom cultivation device can be used for cultivation of edible mushrooms. According to the water mist spraying device, the water mist spraying amount can be dynamically adjusted, meanwhile, the air flow can be synchronously adjusted, water mist is evenly distributed in the whole planting cavity, in addition, redundant water mist and residual water mist after cultivation are reused and flow to the planting tray again through the water permeable plate in the middle of the planting cavity, and the utilization efficiency of water resources is improved.

The invention discloses a drought and flood saline-alkali soil ecological comprehensive improvement method based on an Internet of Things dynamic monitoring technology, which is used in the field of vegetation comprehensive improvement and comprises the following steps: dividing a preset ecological environment area; acquiring soil environment parameters in each area in the ecological environment by using detection equipment, and uploading the encrypted soil environment parameters to a cloud server of a cloud platform; the encrypted soil environment parameters are decrypted and analyzed; deploying an unmanned aerial vehicle for regional patrol; and according to a soil environment parameter evaluation result, an environment change rule and optimization information of equipment layout, combining real-time data of unmanned aerial vehicle patrol, comprehensively analyzing a change trend of a soil environment, and generating a comprehensive improvement plan. According to the method, the soil environment parameter evaluation result, the environment change rule and the equipment layout optimization information are combined, the change trend of the soil environment is obtained, and the soil improvement plan with high pertinence and predictable implementation effect is formulated.

The invention discloses a blast hole parameter adaptive tunnel smooth surface effect prediction method, which comprises the steps of constructing a smooth surface effect prediction model based on a graph neural network; predicting rock mass displacement data by using the smooth surface effect prediction model; and calculating a tunnel smooth surface effect in combination with a prediction result of the rock mass displacement data. According to the method, the adaptive blast hole parameter smooth surface effect rapid prediction model based on the depth map neural network is constructed, smooth surface blasting data are comprehensively extracted by using the map neural network, the accuracy of a deep learning-based calculation method is improved, and the method has important theoretical significance and application value for tunnel blasting construction.

The invention discloses a high-flux microfluidic strain screening device, and relates to the technical field of microfluidics, the high-flux microfluidic strain screening device comprises a microdroplet generation unit and a microdroplet collection unit, the microdroplet generation unit comprises an inner water phase pipe, an oil phase pipe, an outer water phase pipe and a liquid outlet pipe which are coaxially arranged; the micro-droplet collecting unit comprises a collecting plate, a micro-droplet liquid inlet, a micro-droplet collecting cavity and a waste liquid discharging opening are formed in the collecting plate, a plurality of positioning grooves for capturing micro-droplets are formed in the bottom of the micro-droplet collecting cavity, and an extraction channel penetrating to the bottom surface of the collecting plate is formed in the bottom of each positioning groove; one end of the microdroplet liquid outlet of the liquid outlet pipe is inserted into the microdroplet liquid inlet; in order to solve the problems of low flux, easy cross contamination and false positive screening in single strain separation and culture in the prior art, the invention provides a high-flux microfluidic strain screening device capable of quickly separating, culturing and screening a single strain collected in an environment.

The invention discloses a method for predicting the remaining service life of an energy storage battery and a related device. The method comprises the following steps of: establishing maximum discharge energy-circulation circle data of a full life cycle of the battery; sG filtering smoothing is carried out, the maximum discharge energy value of the first circle is used as an initial energy value to convert the maximum discharge energy of the single circle of the battery into SOH, standardization processing is carried out, and a time window sliding method is adopted to divide an obtained standardized data set; constructing a time sequence prediction model based on the LSTM, and training the time sequence prediction model based on the LSTM by using the standardized data set X; the method comprises the following steps: collecting discharge energy data of a to-be-tested battery in an actual operation process in an energy storage system in the (1-n) th circle, inputting the discharge energy data into a pre-trained LSTM time sequence prediction model after standardization processing, outputting predicted SOH values of the (n + 1) th circle to the Nth circle, and establishing a complete battery SOH attenuation curve in combination with the data after standardization processing to obtain the remaining service life of the battery.

The invention discloses an end-to-end calculation method for real-time prediction of an RC component explosion damage effect, and the method comprises the following steps: building an RC component explosion damage effect numerical value solving method, and obtaining RC component explosion damage effect numerical value simulation data; associating the numerical simulation data of the explosion damage effect of the RC component with the explosion working condition and the geometrical shape, the size and the material characteristics of the RC component, and obtaining an RC component explosion damage effect database comprising a training data set and a verification data set; training and optimizing the constructed neural network by using the training data set and the verification data set to obtain an end-to-end calculation model of the explosion damage effect of the RC component; and completing real-time prediction of the explosion damage effect of the target RC component by using the end-to-end calculation model of the explosion damage effect of the RC component. The method for predicting the explosion damage effect of the RC component is simple in process, high in speed and good in seaworthiness.

The invention relates to a driving scoring training system and method based on simulated driving. The driving scoring training system based on simulated driving comprises a simulated driving cabin, a simulated driving simulator, a simulated data server, an information collecting and scoring module and a feedback training data generator, when a driver issues an operation command in the simulated driving cabin, a command signal is transmitted to the simulated driving simulator, and the simulated driving simulator is used for simulating the driving of the driver; the simulation data server generates a virtual model of the vehicle and the road condition, and feelings information is fed back to a driver; the information collecting and scoring module scores the complete driving behaviors and gives driving training suggestions; and the feedback training data generator generates a new training simulation model for training the driver according to the driving training suggestions and by combining the existing training model route data and traffic data. According to the driving scoring training system based on simulated driving, the driving skill of a driver can be improved in a targeted manner, the road safety risk is reduced, and the problem is prevented from occurring again.

The embodiment of the invention discloses an application scene simulation interaction method and device for ship visual signals. The method comprises the following steps: constructing an application scene library according to use specifications and display requirements of ship visual signals; obtaining work information of a first target person to be trained at present, determining a selection weight of each scene type according to the work information, and randomly selecting a target number of first target application scenes in an application scene library according to the selection weight; and displaying the first target application scenes in sequence from low to high according to the scene difficulty coefficients, obtaining simulation interaction data, and generating a scene training result according to the simulation interaction data. According to the embodiment of the invention, the selection tendencies of different first target application scenes can be adjusted according to the selection weights determined according to the work information of the first target personnel, so that the finally displayed first target application scene is ensured to accord with the ability level of each first target personnel, and the training effect is better; and people with any ability level can be effectively trained.

The invention discloses a semi-supervised optical flow prediction method for cell migration path tracking. The technical problem that a model obtained by directly using existing synthetic optical flow data for training is difficult to be directly applied to the field of cell tracking is solved. Firstly, in view of the limitation that an existing convolution kernel is fixed under the condition of processing non-rigid motion of cells, an anisotropic convolution block of a novel convolution mode is provided, and details of motion edges in a cell image are accurately modeled; and then, in order to solve the problem that movement of the optical flow is difficult to distinguish due to dense textures in a cell image sequence, a multi-dimensional feature fusion mechanism is built, and the mechanism integrates multi-dimensional feature relationships in the horizontal direction, the vertical direction and the channel direction so as to ensure that the model can capture dynamic features in the cell image more comprehensively and accurately. Secondly, aiming at the problem of cell image label missing, a semi-supervised learning method is explored so as to reduce dependence on accurate labels.

The invention provides a deep learning irradiance numerical mode prediction correction method and system introducing clustering, and the method comprises the steps: enabling information sequences under the same generalized weather type to be classified into the same cluster through introducing a clustering algorithm; the variational mode decomposition algorithm can decompose historical data time sequence signals into various intrinsic mode functions, redundant information is removed, and effective features are obtained; the feature learning task is optimized, so that deep learning model training and error correction can be more effectively carried out subsequently.

The invention relates to a training method and device of a single-agent reinforcement learning model based on visual representation, electronic equipment, a storage medium and a computer program product, and the single-agent reinforcement learning model comprises an online state encoder, an action encoder, a reinforcement learning network and an auxiliary task network. The auxiliary task network comprises a state prediction model, and according to state information and action information obtained by a target agent based on an observation image of the target agent in the current time period and reward information of the current time period, the state prediction model is used for predicting the state information and action information of the target agent from the perspective of visual representation through the auxiliary task network. The state characterization and the action characterization of the target agent are learned, the optimal decision action is selected for the target agent through the reinforcement learning network, and the time sequence information of the time period in reinforcement learning is fully utilized. Performance and sample efficiency improvement of a single agent in a challenged complex continuous control task with an image as state input can be realized.

A variational autoencoder (VAE) neural network system, comprising: an encoder neural network for encoding input data items to define a posterior distribution of a set of latent variables; and a decoder neural network for generating output data items representing values ​​of the set of latent variables sampled from the posterior distribution. The system is configured for training with an objective function that includes a term that depends on the difference between the posterior distribution and the prior distribution. The prior distribution and the posterior distribution are arranged so that they cannot match each other. The VAE system can be used to compress and decompress data.

The present invention discloses a method for detecting tumor multi-genes based on hyperspectral images, which belongs to the technical field of mutation sample detection, and specifically includes: obtaining tumor tissue sections of different carcinogenic driver gene mutation states after H&E staining; photographing tumor tissue sections by a hyperspectral camera, obtaining hyperspectral images of tumor samples and preprocessing; building a global feature extractor based on S3Anet; building a tumor multi-gene detection model and using multi-task learning training; collecting tumor samples to be detected, and obtaining hyperspectral data of the samples to be detected by a hyperspectral camera; inputting the hyperspectral data into a trained S3ANet model for feature extraction, and then inputting it into a trained multi-gene detection model to obtain the gene detection results of the tumor samples. The beneficial effects of the present invention are: the present invention effectively reduces the dependence on traditional pathological diagnosis experience and gene sequencing, significantly improves the accuracy, reliability and generalization ability of detection, and reduces the detection cost.

The invention discloses a sound signal noise reduction method and system based on feature fusion. The method comprises the following steps: S1, collecting an actually measured noise signal and constructing a simulated target signal; s2, respectively constructing feature vectors of the actually measured noise signal and the simulated target signal based on features of the actually measured noise signal and the simulated target signal; s3, based on the feature vectors of the actually measured noise signal and the simulated target signal, constructing a fusion feature vector of the simulated sound signal; s4, inputting the fusion feature vector of the simulated sound signal into a multilayer perceptron for training to obtain a trained multilayer perceptron model; and S5, carrying out transfer learning on the trained multi-layer perceptron model to realize actually measured sound signal denoising. Effective training of the multi-layer perceptron is realized by adopting actually measured sound signals and simulated target signals; through transfer learning of a multi-layer perceptron model, under the condition that an actually measured target signal cannot be obtained, noise reduction processing of the actually measured sound signal is realized.

The present invention relates to a blood pressure measurement method, system, device and medium, wherein the method comprises: collecting SCG signals and blood pressure data of a human body; performing three-level noise reduction processing on the collected SCG signals, including: performing preliminary noise reduction on the SCG signals by an ICEEMDAN method to obtain a first signal; performing further noise reduction on the first signal by an RLS adaptive filter to obtain a second signal; performing three noise reduction on the second signal by a soft threshold method to obtain a noise-reduced SCG signal; training a pre-set deep learning model using the blood pressure data and the noise-reduced SCG signal to obtain a trained deep learning model; and measuring the blood pressure of a subject to be measured by using the trained deep learning model. The present invention can accurately measure human blood pressure.

The invention discloses a blasting construction real-time optimization method for a steel frame structure, and the method comprises the steps: constructing a component damage effect prediction model and a steel frame structure mechanical response prediction model based on a graph neural network; predicting the damage effect of the component in the steel frame structure under the blasting impact by using the component damage effect prediction model, and judging a component damage mode according to a component damage judgment standard; predicting the mechanical response of the steel frame structure under the blasting impact by using a mechanical response prediction model of the steel frame structure in combination with the judgment result of the component damage mode; and in combination with a prediction result of the mechanical response prediction model of the steel frame structure, an optimal blasting construction scheme of the steel frame structure is obtained based on a genetic algorithm. The method is convenient to use, efficient in calculation and good in robustness.

The present invention relates to the field of path planning technology, and in particular to a dual-objective path planning method based on deep reinforcement learning, comprising the following steps: S1, representing the state of a road network, and constructing a reward function r, to obtain a dual-objective deep reinforcement learning path planning model to be trained; the state includes a coordinate state, a distance state, and a cu state; S2, training the dual-objective path planning model to be trained, so that it learns an optimal strategy π that can obtain a maximum cumulative reward * , obtain the trained dual-objective path planning model; the cumulative reward is the sum of the immediate reward given by the reward function and the additional future reward estimated by the neural network; S3, use the trained dual-objective path planning model to perform dual-objective path planning. The present invention has the ability to minimize path distance and the versatility to optimize losses and utilities in different scenarios.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training an action selection system using reinforcement learning techniques. In one aspect, a method includes, at each of a plurality of iterations: obtaining a batch of experiences, each experience tuple including: a first observation, an action, a second observation, and a reward; for each experience tuple, determining a state value for the second observation, including: processing the first observation using a policy neural network to generate an action score for each action in a set of possible actions; sampling a plurality of actions from the set of possible actions according to the action scores; processing the second observation using a Q neural network to generate a Q value for each sampled action; and determining the state value for the second observation; and using the state value to determine an update to current values of Q neural network parameters.

The invention discloses a face recognition model training method, device and equipment, which are used for improving the recognition accuracy of a face recognition model. The method comprises the following steps: training an initial face recognition model by using a labeled data set; performing image feature extraction and identification prediction on unlabeled samples in the unlabeled data set by using the initial face identification model to obtain unlabeled sample features and a prediction result, the unlabeled samples being unlabeled face images; according to the label-free sample features and the prediction result, screening a pseudo label sample with the confidence coefficient meeting the condition from the label-free data set, and generating a pseudo label corresponding to the pseudo label sample; training an initial face recognition model by using the labeled data set and a pseudo-label data set, wherein the pseudo-label data set comprises pseudo-label samples and corresponding pseudo-labels;

The present application relates to the technical field of organoid culture, and specifically discloses a culture medium for musk gland organoids and its preparation method and application. The present application discloses a culture medium, which uses DMEM / F12 as the basic culture medium; contains the following concentrations of components as functional additives: albumin, B‑27, HEPES, glutamine, N‑acetyl‑L‑cysteine, niacinamide, growth factor, Wnt3a, A83‑01, Forskolin, penicillin, streptomycin, hydrocortisone, insulin; the growth factor is EGF and bFGF with a weight ratio of 350: (11‑21). The present application also discloses a method for preparing the above-mentioned culture medium and its application in musk gland organoid culture. The culture medium of the present application can be used to obtain musk gland organoids with stem cell characteristics, and the cell activity is high.

The present application provides an intelligent online feeder terminal fault identification method, which first collects power parameters, switch states and environmental data of each feeder terminal, converts these data into feature vectors through a feature extraction algorithm, and then uses a pre-trained abnormal probability recognition model to generate an abnormal probability value and calculate its confidence interval. If the lower limit of the confidence interval exceeds a preset threshold, the terminal is judged to be abnormal. For the abnormal terminal, the method determines its adjacent terminals according to the power grid topology structure, and then uses a pre-trained fault classification model to classify the feature vectors of the abnormal terminal and the adjacent terminal respectively to obtain their respective fault classification data. Finally, the method comprehensively considers the fault classification data of the abnormal terminal and the adjacent terminal, generates a comprehensive fault classification result of the abnormal terminal, and realizes intelligent online identification of feeder terminal faults, which helps to improve the accuracy and efficiency of power grid fault diagnosis.

The invention discloses a serial crystallography experimental device based on a conveyor belt technology. The device comprises an angle measuring head end, a bottom film end and a crystallization plate storage disc, an angle measuring head groove is formed in the bottom end of the angle measuring head end and is used for being matched with an angle measuring head base on a synchrotron radiation beam line station; the top end of the angle measuring head end is provided with two angle measuring head end supporting rods used for being matched with the bearing short rods, and an angle measuring head end waste liquid tank used for collecting waste liquid irradiated by X rays. The bottom film end is assembled with the crystallization plate storage disc and is used for supporting and turning over the crystallization plate storage disc; in an experiment, a crystallization plate storage disc provided with a sitting drop in-situ crystallization plate is assembled on the bottom film end, and a low background scattering film on the back surface of the sitting drop in-situ crystallization plate is connected with a bearing short rod through an adhesive film to form a conveyor belt film; when the bearing short rod is driven by the motor to convey a sample to the angle measuring head end, the conveying belt film is torn down, diffraction data collection is carried out, and waste sample liquid drops are extruded out, flow down along with the angle measuring head end supporting rod and are converged into the waste liquid tank.

The present invention discloses a method and system for reducing noise of acoustic signals based on feature fusion, the method comprising: S1: collecting measured noise signals and constructing simulated target signals; S2: constructing feature vectors of the measured noise signals and the simulated target signals based on the features of the measured noise signals and the simulated target signals respectively; S3: constructing a fused feature vector of the simulated acoustic signal based on the feature vectors of the measured noise signals and the simulated target signals; S4: inputting the fused feature vector of the simulated acoustic signal into a multilayer perceptron for training to obtain a trained multilayer perceptron model; S5: performing transfer learning on the trained multilayer perceptron model to achieve denoising of the measured acoustic signals. The present invention adopts measured acoustic signals and simulated target signals to achieve effective training of the multilayer perceptron; through transfer learning of the multilayer perceptron model, when the measured target signal cannot be obtained, the noise reduction processing of the measured acoustic signal is achieved.