81results about How to "Guaranteed training effect" patented technology

The invention provides a slope stability dynamic prediction and geological parameter inversion method, and the method comprises the steps: building a slope stability simulation model through a COMSOL multi-physics field simulation platform, and generating simulation data in combination with an orthogonal experiment design and automatic parameter optimization technology; then, training and optimizing the simulation data based on a variational auto-encoder and a generative adversarial network to obtain enhanced simulation data; and performing geological parameter inversion based on SBAS-InSAR ground deformation data and the enhanced simulation data, generating a safety coefficient distribution map in the region by using geological parameters obtained by inversion, and fusing the safety coefficient distribution map with a geological disaster potential map to evaluate disaster risks of different regions. According to the method, the timeliness and the real-time performance of a prediction result are ensured, and a scientific basis is provided for engineering design, disaster monitoring and decision making.

The invention belongs to the technical field of data analysis, and particularly relates to an information analysis method for bidding documents. The information analysis method comprises the following steps: S1, obtaining a newly added bidding and tendering file, and sending the file into a preprocessing module; s2, the preprocessing module outputs corresponding preprocessing data to a classification module; s3, the classification module divides the content in the current preprocessed data into valid tables, core blocks and independent blocks, and then the valid tables, the core blocks and the independent blocks are sent into a table analysis module, a core analysis module and an independent analysis module respectively; and S4, the table analysis module, the core analysis module and the independent analysis module respectively output the first analysis data, the second analysis data and the third analysis data to an integration module for integration to obtain core information and output the core information. According to the method, the core information of various bidding and tendering files can be accurately and efficiently analyzed.

The invention discloses a medical data processing method and apparatus, and an electronic device. The method comprises the steps of constructing a plurality of single-mode prediction models and a plurality of single-mode training data sets in one-to-one correspondence with the single-mode prediction models; based on the single-mode training data set, training a single-mode feature extractor in the corresponding single-mode prediction model; constructing a multi-modal prediction model and a multi-modal training data set; and based on the multi-modal training data set, training a multi-modal classifier in the multi-modal prediction model to obtain a trained multi-modal prediction model. The trained multi-modal prediction model can perform accurate feature representation on the medical data of the multiple modals, and on the basis, the risk of the coronary heart disease can be accurately predicted based on the medical data of the single modal or the medical data of the multiple modals through the multi-modal prediction model.

The invention relates to a video key frame obtaining method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a plurality of candidate image frames corresponding to a to-be-processed video, carrying out the preprocessing of each candidate image frame, inputting the preprocessed candidate image frame into a pre-constructed image feature extraction model, obtaining a feature vector corresponding to each candidate image frame, and obtaining a feature vector corresponding to each candidate image frame; performing clustering processing on the feature vector corresponding to each candidate image frame through the target clustering parameter based on a preset clustering algorithm to obtain at least one target clustering result, and performing calculation based on the feature vector corresponding to each candidate image frame and the center vector corresponding to the target clustering result to obtain a target clustering result; and determining a target image frame of the to-be-processed video from the plurality of candidate image frames based on a calculation result. By adopting the technical scheme, the image frames with large scene and element differences can be screened from the continuous image frames, and the screened image frames are labeled and applied to the training of the automatic driving perception algorithm, so that the training efficiency can be greatly improved.

The invention discloses a video anomaly detection method based on a multi-modal condition diffusion model, and belongs to the technical field of video anomaly detection. The method comprises the following steps: preprocessing a video to generate a video space-time cube and an optical flow space-time cube; constructing a diffusion model, and obtaining an abnormal score; performing weighted calculation on the reconstruction anomaly score and the prediction anomaly score; and evaluating the obtained final anomaly score to complete video anomaly detection. According to the invention, through a multi-task combination mode of optical flow reconstruction and video frame prediction, abnormity is easier to distinguish; multi-modal data are fused, feature information of more dimensions is captured, and the accuracy of anomaly detection is improved; the model is more stable when processing noise and complex backgrounds, and the stability of anomaly detection is improved; the reconstructed optical flow is used as a condition for guiding future frame prediction, the prediction error between an abnormal prediction frame and an original frame is expanded, the prediction error between a normal prediction frame and the original frame is reduced, and the training effect and performance are ensured.

The invention relates to the technical field of radar weak target detection, in particular to a radar weak target detection method and device based on deep learning. According to the technical scheme, the method comprises the following steps: data preprocessing and enhancement: receiving an original echo signal of a radar, generating three-dimensional time frequency-wavelet coefficient spatial data by combining time frequency analysis with wavelet transform, and expanding a data set in multiple modes; and feature extraction network construction: constructing a 3D-CNN and LSTM fused network, wherein the 3D-CNN has a convolution operation with a variable convolution kernel size. According to the method, the data are richer and more comprehensive through time-frequency analysis in combination with wavelet transform and data enhancement, and multi-dimensional features can be accurately captured through 3D-CNN and LSTM fusion and an adaptive convolution kernel; and the training effect is ensured through the adaptive learning rate and the early stop method, and the accuracy and adaptability of weak target detection are jointly improved.

The invention relates to the technical field of computers, and discloses a dialogue simulation method and device, equipment, a storage medium and a program product. The method comprises the steps that dialogue course information is acquired, the dialogue course information is displayed in a course display area of a dialogue interface, and the dialogue course information comprises at least one dialogue course; in response to a selection operation for a dialogue course in the dialogue course information, displaying a simulated dialogue of the dialogue course in a dialogue display area of the dialogue interface, the simulated dialogue corresponding to a simulated reply; responding to a reply operation for the simulated dialogue, and obtaining reply content; and generating a dialogue simulation result based on the matching degree of the reply content and the simulation reply, and displaying the dialogue simulation result in a result display area of the dialogue interface. By implementing the technical scheme of the invention, diversified dialogue simulation training and real-time feedback of the reply content are realized, the course training efficiency is improved, and the course training effect is ensured.

The invention relates to a multi-unmanned aerial vehicle area coverage path rapid planning method, and relates to the field of non-electrical variable control or adjustment systems. The method comprises the following steps: using an intelligent path planning model to plan three-dimensional coordinate data corresponding to each space point of a flight area covered by a certain flight path used by a current unmanned aerial vehicle group for completing a target flight path; the flight duration corresponding to a certain flight path is intelligently predicted according to various route data and multiple flight configuration information; and taking the flight path corresponding to the minimum value in the flight durations as a planned flight path. According to the invention, the technical problems that the minimization of the total flight time and the obstacle constraint are difficult to consider, and the path planning operation cost and the time cost are high are solved; and an intelligent path planning model is adopted to traverse flight durations corresponding to flight paths simulated by the current unmanned aerial vehicle group for completing the target flight line based on multiple basic data, so that the technical problem is solved.

The utility model relates to the field of training devices, and discloses a dorsal muscle training device which comprises a balance weight base, a supporting frame is fixedly arranged at the top of the balance weight base, a mounting groove is formed in the top of the balance weight base, a hydraulic cylinder is fixedly arranged in the mounting groove, and a platform is fixedly connected to a hydraulic rod of the hydraulic cylinder. The dorsal muscle training mechanism is arranged on the supporting frame, the dorsal muscle training mechanism comprises a supporting piece, a sliding plate, a rope body, a connecting piece, a tension spring, a handle, a transverse frame, a vertical frame, a display screen, a counting sensor and a counter, two rectangular grooves are formed in the side face of the supporting frame, and the supporting piece is fixedly installed in the rectangular grooves; the sliding plate is arranged on the outer side of the supporting piece in a sleeving mode. The dorsal muscle training device has the following advantages and effects that the training quantity can be checked through the display screen, automatic counting processing can be conveniently carried out, trainees are prompted according to the counting condition, and the dorsal muscle training effect can be guaranteed.

The invention discloses an unmanned aerial vehicle model pruning method and system based on federal reinforcement learning, and relates to the federal learning and reinforcement learning technology, and the method comprises the steps that in a pruning round, a server uses a reinforced generalization pruning method (RGP) to prune a global model, and a pruned model is obtained; the server performs screening to obtain clients participating in pre-training of the current round, and issues the pruned model or the global model to the clients of the current round; the client receives the pruned model or global model sent by the server, and performs fine tuning by using local data; and the client transmits the fine-tuned model to the server, and a sparse model aggregation method (FedSA) is used in the server for model aggregation to obtain a new global model. According to the method, the training effect of federal learning at the edge end with limited resources can be optimized, and the complexity of the training process is reduced while the federal model effect is kept.

The rooting culture device comprises a bearing mechanism, an LED lamp is arranged above the bearing mechanism, bases are installed on the two sides of the LED lamp, the bases are placed on the two sides of the bearing mechanism, and a covering and sealing mechanism is movably connected to the upper portion of a rear panel body at the top of the bearing mechanism. A side blocking mechanism is slidably connected to the right side of the bearing mechanism, and a plurality of supporting mechanisms are evenly arranged in the middle of the interior of the bearing mechanism in parallel at equal height. A living space is provided for dendrobium candidum through the bearing mechanism, a humid environment is provided for the dendrobium candidum by laying the moss, the white light LED lamps provide full-spectrum light similar to natural sunlight, growth and development of the dendrobium candidum are facilitated, the dendrobium candidum supporting rods are limited through the multiple supporting mechanisms, and the dendrobium candidum supporting rods are limited through the multiple supporting mechanisms. Therefore, the culture efficiency of the dendrobium officinale is ensured.

The present invention relates to a video surveillance and fire alarm linkage system and method, and relates to the general field of image data processing or generation. The system includes: a fire analysis mechanism, which is used to input the floor area of ​​a fire monitoring area, multiple foreground information of a panoramic image, and various content data of the panoramic image into an intelligent fire analysis model to obtain the interval time from the image shooting time to the fire occurrence time; a wireless transmission mechanism, which is used to wirelessly transmit the interval time to a fire management service network element. The present invention also relates to a video surveillance and fire alarm linkage method. Through the present invention, in the face of the technical problem that the time of fire occurrence cannot be predicted based on current visualization data, the specific time of subsequent fire occurrence is intelligently judged by using an intelligent fire analysis model based on the floor area of ​​the fire monitoring area and multiple targeted screened visualization data of the panoramic picture of the fire monitoring area at the current moment.

The embodiment of the invention provides a key phrase generation model training method and device and a key phrase generation method and device. The training method comprises the following steps: acquiring a training document and a plurality of key phrase true values corresponding to the training document; distributing a plurality of key phrase true values to a plurality of control codes based on an optimal transmission theory; inputting the training document into a generator to enable the generator to output a plurality of candidate key phrases; inputting the plurality of candidate key phrases into a selector to enable the selector to output a decision label which is used for indicating that the corresponding candidate key phrases are reserved or discarded; and constructing a first loss function corresponding to the generator and a second loss function corresponding to the selector, and adjusting and optimizing the generator and the selector according to the first loss function and the second loss function to obtain a target key phrase generation model. According to the technical scheme, the accuracy and recall rate of key phrase generation can be considered, and the key phrase generation quality is guaranteed.

The utility model discloses a multifunctional shaking table for microbial culture, which is characterized in that the end part of the output shaft of a driving motor is fixedly connected with a driving round wheel, the outer side of the driving round wheel is provided with a driving frame, the middle ends of the two side parts of the driving frame are fixedly connected with sliding rods, and the top of a table plate is fixedly provided with limiting frames at the two sides of the driving frame; a reciprocating driving structure is formed by the driving motor, the driving round wheel and the driving frame, the sliding rod is conveniently driven by the reciprocating driving frame to slide along the limiting frame in a limiting and guiding mode, the sliding rod can slide along the limiting frame in a matched mode, and the sliding bottom plates are fixedly connected to the top positions of the two sides of the sliding rod. A sliding bottom plate is driven by a sliding rod to move back and forth, so that the mounting box and the culture shaking table are driven to move back and forth, stably and horizontally, microorganisms in the culture shaking table can make more sufficient and uniform contact with a culture solution, and the microorganism culture effect is guaranteed.

The invention relates to the technical field of machine learning, in particular to a federal learning acceleration method and system for industry user water consumption prediction. According to the federal learning acceleration method for industry user water consumption prediction, in the global training process, clustering is performed according to the local training speed of edge servers at intervals of multiple global training times, and multiple edge server clusters are formed; the edge servers in the edge server cluster are globally aggregated to update the local model. According to the method, the edge servers are clustered, and intra-cluster model aggregation is carried out by taking the edge server cluster as a unit, so that the combination of synchronous training and asynchronous training is realized. In the global training process, dynamic combination of edge server clusters is realized according to local training time interval clustering, so that the correlation of all edge servers in the global training process is ensured, and the model training performance is ensured; the clusters are mutually independent, and the aggregation in the clusters greatly saves the model training time compared with the aggregation of all edge servers.

The utility model belongs to the technical field of fitness equipment, and relates to a winding device and the fitness equipment. The winding device comprises a reel, a rope and a guide sleeve. The reel comprises a winding section and a guide section which are arranged in the axial direction of the reel, and a spiral guide groove is formed in the outer wall of the guide section. The guide sleeve sleeves the guide section and is in threaded connection with the guide section; the rope is wound on the winding section, one end of the rope is connected to the reel, and the other end of the rope is separated from the reel to be connected with a handle assembly for training; when the reel rotates back and forth under the action of external force, the guide section and the guide sleeve are mutually screwed, so that the reel does reciprocating motion relative to the guide sleeve in the axial direction of the reel, and the rope is retracted and pulled. The rope winding device can avoid the phenomenon of line folding in the rope winding and pulling process, and ensures that the rope is stably wound out, so that obvious pause feeling is avoided in the rope winding and pulling process, the continuity of training actions of a user is ensured, and the training effect is ensured.

The invention relates to medical image data processing, in particular to a medical image report splitting method for large model training, and the method comprises the steps: reading a to-be-split medical image report, and carrying out the sentence segmentation of the medical image report, and obtaining a combination of segmented sentences seen for examination and a combination of segmented sentences of examination impression; extracting word vectors and mask vectors corresponding to each check-seen clause in the check-seen clause combination and each check-impression clause in the check-impression clause combination; inputting the word vector and the mask vector corresponding to each check-seen clause in the check-seen clause combination into a pre-trained check-seen classification model to obtain a check-seen classification result corresponding to each check-seen clause; inputting the word vector and the mask vector corresponding to each check impression clause in the check impression clause combination into a pre-trained check impression classification model to obtain a check impression classification result; according to the technical scheme provided by the invention, the defect that the examination impression in the medical image report is difficult to accurately split according to categories can be overcome.

The invention discloses an unmanned aerial vehicle ground moving target training system and method, and belongs to the technical field of moving target training, and the method comprises the following steps: receiving an input site topographic map and vehicle information, and carrying out random driving and storage; drawing a corresponding vehicle main body path diagram for each vehicle number, and generating a vehicle random path diagram according to the vehicle driving range; sending the vehicle main body path diagram to the vehicle with the corresponding vehicle number; the vehicle detects whether the vehicle is locked or not; if it is detected that the vehicle is locked, driving is carried out according to the vehicle random path diagram.The method has the advantages that the vehicle can be driven randomly after being locked, the training difficulty is effectively increased, and actual combat simulation is more real.

The invention provides an image data processing method and device, electronic equipment and a storage medium, and relates to the technical field of image processing and the field of financial science and technology. The method is applied to a security platform end, and comprises the following steps: storing a sample image to an archiving database, and generating an image storage path; generating image training data according to the image structured data and the image storage path, and storing the image training data to a distributed database; obtaining image training data from a distributed database according to the model training request, extracting a sample image from an archiving database according to an image storage path in the image training data, and training a preset initial image processing model according to the image training data and the sample image in a safe execution space to obtain a target image processing model; and obtaining a target image, and performing image processing on the target image through the target image processing model. According to the embodiment of the invention, model training can be carried out on the premise of ensuring data security, and the model training effect is ensured.

The utility model discloses a culture bottle for indoor soil culture, and relates to the technical field of soil sample culture. A 150 cylindrical brown glass wide-mouth bottle is selected as the culture bottle, the brown glass can prevent light pollution, the culture effect and quality can be further guaranteed, the cylindrical culture bottle can be arranged in the culture box in the maximum volume mode, the bottom of each silicone tube extends into the bottom of the culture bottle, and therefore the culture effect and quality are improved. Through the arrangement, gas in the culture bottle can be fully exchanged; by arranging the three-way valve, the installation stability of the silica gel tube can be guaranteed, nitrogen blowing can be carried out on the interior of the silica gel tube to guarantee the CO2 background value of a culture bottle, repeated uncovering is not needed, sample pollution is avoided, and continuous sampling measurement of a soil sample culture experiment is facilitated; the purpose of rapid sampling can be achieved through the injector, and the cultivation effect and quality are guaranteed.

A distributed model training optimization method, system, medium and product relate to the field of model training, by adopting the above technical scheme, the training optimization system divides a terminal device into a plurality of device groups and sets a relay node in each device group, when a preset synchronization condition is satisfied, there is no need to send the information to the terminal device one by one, and the synchronization efficiency is improved. And the model parameters only need to be sent to the relay node, so that hierarchical transmission of the model parameters is realized. After the parameter difference value is calculated, the training optimization system screens the to-be-updated parameters, meanwhile, the high-frequency synchronization parameter set and the low-frequency synchronization parameter set are determined according to the synchronization frequency of the to-be-updated parameters so that different to-be-updated parameters can be sent at different moments, and due to the differentiated parameter synchronization strategy, important parameters can be updated in time, and the updating speed of the to-be-updated parameters is improved. And the performance bottleneck caused by frequently transmitting all parameters is avoided, and the network communication overhead is remarkably reduced, so that the overall efficiency of distributed training is improved.

The utility model belongs to the technical field of fitness equipment, and relates to a winding device and the fitness equipment. The winding device comprises a winding wheel, a rope and a transmission assembly. A spiral groove is concavely formed in the outer wall of the reel; the transmission assembly comprises a guide rail, a guide sleeve and a shifting piece, the guide rail is arranged outside the reel in parallel, the guide sleeve sleeves the guide rail and is in sliding connection with the guide rail, the shifting piece is convexly arranged outside the guide sleeve, and the head of the shifting piece extends into the spiral groove; the rope is wound in the spiral groove, one end of the rope is connected to the reel, and the other end of the rope is separated from the reel and connected with a handle assembly used for training. When the reel rotates back and forth under the action of external force, the shifting piece is driven by the reel to drive the guide sleeve to move back and forth along the guide rail, and the rope is pulled and retracted. The rope folding phenomenon in the rope pulling process can be avoided, it is ensured that rope outgoing is stable, and therefore the obvious pause feeling in the rope pulling process is avoided, the continuity of training actions of a user is ensured, and the training effect is ensured.

The utility model discloses a cell culture bottle rack relates to cell culture device technical field. The cell culture bottle rack includes rack base, support platform, clamping unit, limit block, clamping block and locking unit, and support platform slidingly connects in the top of rack base. The cell culture bottle rack passes through the design of support platform, clamping unit and limit block, makes cell culture bottle convenient for clamping between two sets of limit blocks, keeps clamping stable, and cell culture effect is guaranteed, and simultaneously, through the design of limit block, clamping block and locking unit, makes limit block convenient for dismounting and replacement model, to adapt to multiple models cell culture bottle, ensures that the contact area between cell culture bottle and limit block is big during clamping process, and clamping effect is good, and safety is good, satisfies use demand.

The embodiment of the invention provides a data acquisition method and device, computer equipment and a readable storage medium. The method comprises the steps that data demand information is acquired; the data demand information carries data feature information; determining a plurality of target training sample data from the initial training sample data according to the data feature information and label information corresponding to the initial training sample data; querying a preset database, and obtaining target storage corresponding to each piece of target training sample data; and sending a data acquisition request to each target storage to acquire target training sample data. According to the data acquisition mode provided by the embodiment of the invention, different training data can be integrated, and the appropriate training sample data can be automatically requested and acquired based on the demand of the user for the training data, so that the effect of subsequent model training is ensured.

The invention relates to the technical field of lower limb training, and discloses a Parkinson's patient lower limb training aid, which comprises a thigh seat and a shank seat, which are respectively fixed on the rear sides of thighs and shanks; the upper connecting rod is arranged at the bottom of the thigh seat, and the bottom end is fixedly connected with a fixing plate; the top end of the lower supporting rod is fixedly connected with a rotating disc; the fixing plate and the rotating disc are rotationally connected, a resistance mechanism is arranged at the joint, and the smaller the included angle between the upper connecting rod and the lower supporting rod is, the larger the resistance provided by the resistance mechanism is. According to the lower limb training assist device for the Parkinson patient, through dynamic resistance provided by the arc-shaped spring, the backward bending phenomenon of shanks of the patient in high leg lifting training can be effectively limited, training actions are standardized, it is guaranteed that target muscle groups are fully activated, and therefore the training effect is guaranteed, and the training efficiency is improved.

The invention relates to an air conditioner energy-saving control method and system of a digital twinborn model, and belongs to the field of air conditioning. The method comprises the steps that the temperature adjusting range of a target refrigeration air conditioner is evenly divided to obtain a plurality of reference set temperatures, inputting each reference set temperature into a digital twin model to obtain predicted consumption power consumption and predicted consumption duration corresponding to the reference set temperature; the reference set temperature corresponding to the predicted power consumption with the minimum value serves as the preferential energy-saving set temperature; the technical problem that in the prior art, an air conditioner is difficult to obtain the optimal initial refrigeration temperature is solved. A traversal type processing mechanism is used, and a digital twinborn model designed by a customized structure is adopted to intelligently predict a plurality of consumed powers and a plurality of continuous durations when the refrigeration air conditioner reaches the same expected refrigeration temperature at different initial set temperatures, so that the technical problem is solved.

The present disclosure provides a method and device for improving GAIN to repair random missing of multi-sensor data. The improved GAIN includes a generator and a discriminator. The generator and the discriminator both include an encoder-decoder architecture composed of a convolutional neural network, and a self-attention mechanism is integrated in the generator to make the network focus more on important features and enhance the ability of the network to extract global features. Both the generator and the discriminator are provided with jump connections, which can alleviate the problem of gradient disappearance and improve the utilization rate of features. The method trains the improved GAIN by using an incomplete data matrix, and the data corresponding to each sensor in the incomplete data matrix are randomly partially missing. Thus, the sensor data actually collected is repaired according to the trained generator, which can ensure the repair effect of the missing data of all sensors.

The invention relates to an image generation method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring an initial image and a target description text; performing feature extraction on the target description text to obtain overall description features and object description features; feature extraction is carried out on the initial image to obtain a plurality of image feature layers, and the plurality of image feature layers comprise a semantic feature layer and a relation feature layer; sequentially adjusting the plurality of image feature layers according to the object description features and the overall description features to obtain a plurality of adjusted image feature layers; the multiple adjusted image feature layers are fused, an adjusted image is generated, the adjusted image is used for generating a training image set, and the training image set is used for training an image recognition model. By adopting the method, the efficiency and the image diversity can be improved.

The present invention discloses a method and device for mining recommendation information, and relates to the field of computer technology. A specific implementation of the method includes: extracting attribute elements of a set attribute type from text description information of a target object to obtain scene attribute information of the target object; combining the attribute elements of the scene attribute information according to a set combination strategy to generate scene combination information including event attribute elements; generating recommendation information of the target object according to the attribute elements of the scene combination information and a set recommendation template. This implementation combines the attributes of the target object itself with the user's intention, which can not only highlight the target object, but also meet the user's needs, improve the recommendation effect, and enhance the user experience.

The present invention discloses an occlusion-aware seamless segmentation method and system for pseudo-label generation and passive domain scene adaptation. The method further improves the labeling accuracy of pseudo-labels by threshold screening and data volume comparison for conventional instance-level pseudo-labels. At the same time, the generated pseudo-labels are used to design a weighted loss guided by the uncertainty region of the instance-level prediction branch in the occlusion-aware seamless segmentation task, thereby improving the accuracy of the segmentation model. Combined with the screening of low-quality pseudo-labels, a non-modal guided instance mixing strategy is proposed, thereby further increasing the number of samples available for training. The problem of poor segmentation effect of the final trained model caused by the small number of samples of individual categories is thereby improved, and the passive domain scene adaptation effect is finally improved.