2446 results about "Learning network" patented technology

The invention provides a reinforcement learning-based database adaptive data flow acquisition optimization method and system, and relates to the technical field of databases, and the method comprises the steps of constructing a distributed acquisition proxy network, obtaining database data flow information, and carrying out multi-modal feature decomposition and fusion. And calculating an optimal initial acquisition parameter combination by using the fusion feature vector and combining a genetic algorithm, constructing a double-layer deep Q learning network structure, performing hierarchical training, and generating a hierarchical acquisition strategy model. The model adaptively adjusts a global acquisition strategy and parameters according to a characteristic drift degree and a performance index so as to realize continuous optimization of the acquisition strategy. According to the method, through multi-modal feature fusion, reinforcement learning and a self-adaptive tuning mechanism, the data flow collection efficiency is effectively improved, the collection delay is reduced, and the resource utilization rate is optimized.

The present invention provides an automatic intelligent inspection system for the equipment, facilities and surrounding environment along a railway line, including an unmanned aerial vehicle (UAV), a movable ground base station, a remote server and a working computer, wherein the working computer includes an image processing module that is configured to detect an obtained image by using a deep learning network model trained with a database in advance, obtain any object that is suspected to involve a defect by analysis, and transmit the object for manual review while calculating for the specific geographic coordinate information of the defect, report for early warning after the defect is confirmed by the manual review, inform the operation and maintenance personnel, and store and record the defect. The present invention further provides a method for inspection by using the automatic intelligent inspection system for the equipment, facilities and surrounding environment along a railway.

The invention relates to the technical field of digital factories, and discloses an equipment real-time monitoring and fault prediction method and system based on a digital factory. The method comprises the following steps: carrying out iterative filtering interpolation processing on equipment operation data collected by a multi-modal sensor in a digital factory to obtain smooth time sequence data, and constructing a multi-modal feature matrix based on the smooth time sequence data; inputting the multi-modal feature matrix into a joint training generalized Siamese-like wide learning network for model training to obtain an equipment state monitoring model; executing system-level state evaluation and subsystem-level fault prediction calculation based on the equipment state monitoring model, and generating an equipment operation and maintenance strategy; according to the equipment operation and maintenance strategy, a fault risk heat map is generated, a digital twin visualization system is constructed, and an equipment maintenance scheduling scheme is output, the production efficiency and equipment health management are balanced, and collaborative optimization of a maintenance plan and a production plan is achieved.

The invention provides an anomaly detection and maintenance decision optimization method and system for automatic test equipment, and relates to the technical field of automation, and the method comprises the steps: deploying an edge calculation unit and a multi-level sensor network to collect equipment data, carrying out the fault diagnosis through a deep network model, and constructing a maintenance decision model in combination with equipment operation data. The model is input into a quantum enhancement dual-network system to calculate a reward value, and is used for training a maintenance strategy network under a federal reinforcement learning framework. And generating a maintenance resource configuration scheme through a quantum evolutionary algorithm, and generating a maintenance scheme through optimization of the multi-criterion evaluation model and the self-adaptive multi-target differential evolutionary algorithm. According to the scheme, a high-fidelity digital twin model and a multi-task deep transfer learning network are input for verification and updating, updated data are transmitted to a field control system through an industrial Internet of Things platform, and the updated data are recorded by using a block chain technology for updating a deep network model.

The invention provides a multi-level attribution and recommendation method and system for medical management decisions, and relates to the technical field of medical management decisions, and the method comprises the steps: extracting graph node evolution characteristics through a multi-level medical data knowledge graph and a decision knowledge base by employing a time sequence diagram convolutional network, constructing a decision influence propagation graph through a cross-layer interactive attention network, and carrying out the analysis of the decision influence propagation graph; calculating the contribution degree of each decision node to the medical effect, generating an attribution analysis result, determining a decision influence hierarchical system according to the attribution result, and generating and evaluating candidate decision paths in combination with a multi-objective optimization network, a hierarchical reinforcement learning network and a Monte Carlo search tree network. In conclusion, accurate attribution analysis and multi-objective optimization of the management decision can be realized, the hospital resource allocation efficiency is effectively improved, the operation cost is reduced, scientific decision support is provided for a hospital management layer, and the hospital fine management level and the overall operation efficiency are remarkably enhanced.

The invention discloses a multi-modal diagnosis and treatment data analysis and auxiliary diagnosis method for internet medical treatment, which comprises the following steps: S1, performing feature extraction and fusion on multi-modal diagnosis and treatment data through a nonlinear mapping algorithm based on quantum state optimization; s2, constructing a self-adaptive dynamic relation network, and capturing a complex interaction relation between cross-domain features in the diagnosis and treatment data; s3, fusing implicit features in the sparse medical data with semantic association; s4, optimizing the diagnosis and treatment decision process by applying a causal inference technology; s5, processing anomalies in the diagnosis and treatment data by using a fusion generative adversarial network and sparse tensor decomposition technology; s6, constructing a decentralized encrypted federated learning network to protect the privacy of the patient; s7, predicting a disease evolution path in real time to form dynamic decision support; and S8, performing accelerated analysis on the high-dimensional feature mode in the diagnosis and treatment data in combination with quantum calculation. The method has the advantages of high data fusion capability, high diagnosis precision and high real-time decision support capability.

The invention provides an unmanned aerial vehicle return detection method and system based on AI recognition, and relates to the technical field of image recognition, and the method comprises the steps: extracting multi-scale features through an improved moving window hierarchical converter model, obtaining target attitude information through the combination of cross-modal comparative learning and a dynamic routing capsule network, and obtaining target attitude information; an alignment feature representation is obtained by using a feature enhancement module and an adversarial feature alignment method, the alignment feature representation is input into a graph attention diffusion model to obtain an accurate state representation, and a state evolution trajectory is obtained in combination with an adaptive causal convolutional network and an ordinary differential equation network; posterior distribution is determined through a variational inference network and an autoregression flow model, a state sequence is predicted by using a neural process network and a dynamic point process model, an expert strategy is learned by using a hierarchical generative adversarial imitation learning network and a distributed trust domain evolutionary strategy algorithm, and optimal parameter configuration is obtained; and accurate prediction of the return state of the unmanned aerial vehicle is realized.

The invention relates to the technical field of image processing, and discloses a detection method, system and equipment for intelligent welding of pipe pile splicing and a medium. The method comprises the steps that noise reduction is conducted on a target weld seam image through a filtering algorithm, registration and splicing are conducted on the image through feature matching, and a weld seam unfolding image is obtained; segmenting the weld joint expanded image by using a deep learning network, and extracting a weld joint center line and surface features to obtain weld joint feature data; defect detection and recognition are conducted on the weld joint feature data, hierarchical calculation is conducted through a rule set, and weld joint quality data are obtained; parameter mapping is conducted according to the weld joint quality data, a welding parameter instruction is generated, welding process variables are adjusted, and target welding parameters are obtained; and the welding seam quality data and the target welding parameters are stored, and a quality report is generated through data analysis. The detection accuracy of intelligent welding for pipe pile splicing is improved.

The invention relates to a multi-protocol transmission text data monitoring and warning method and system, and the method comprises the steps: generating a multi-source protocol transmission instance based on dynamic authorization and hardware security verification, and collecting and analyzing text data; a network connection state, a data backlog amount and sensor numerical value content parameters are monitored in real time through multiple threads, and a transmission state and content exception event queue is generated; learning a causal relationship among network congestion, equipment faults and alarm events by using a Bayesian network algorithm, calculating a root cause probability in combination with a dynamic weight distribution strategy, and generating a comprehensive alarm list of priority ranking; on the basis of user feedback data, protocol weights and alarm strategies are adaptively updated, abnormal early warning triggering, data snapshot binding and closed-loop optimization of alarm logs are achieved, and the problems that in a multi-protocol mixed transmission scene, safety adaptability is poor, the monitoring dimension is single, root cause analysis depends on static rules, and strategy updating lags are solved. And the real-time performance, the accuracy and the self-adaptability of data transmission of the industrial Internet of Things are improved.

The invention discloses a reinforcement learning unmanned forklift truck obstacle avoidance scheduling method and system for a dynamic obstacle, relates to the technical field of unmanned driving, and discloses the reinforcement learning unmanned forklift truck obstacle avoidance scheduling method for the dynamic obstacle. According to the method, the system and the system, the system and the system, global path planning and local obstacle avoidance decision making are carried out in combination with the reinforcement learning network, the problems of obstacle avoidance response delay and unreasonable path planning in a dynamic environment are solved, the obstacle avoidance response speed, the path planning rationality and the multi-modal data fusion precision in the dynamic environment are improved, and then the safety and efficiency of unmanned forklift dispatching are improved.

The invention relates to a drug-disease relationship prediction method and system based on a dual-channel fusion knowledge graph. The method comprises the following steps: S1, constructing a biomedical fusion knowledge graph; s2, constructing a drug-disease sub-graph based on the biomedicine fusion knowledge graph; s3, constructing a dual-channel adaptive fusion feature module, and embedding a drug-disease sub-graph; s4, performing enhanced splicing on the drug-disease sub-graphs in the dual-channel fusion knowledge graph pre-embedded network, importing an edge-node iterative updating learning mechanism, training a spliced sub-graph relation perception learning network, and obtaining enhanced sub-graph feature embedding; and S5, utilizing the trained enhanced subgraph features to embed, calculate and output the prediction probability of the drug-disease relationship. According to the method, the semantic representation capability of the knowledge graph and the topological modeling advantages of the graph neural network are integrated, and efficient and accurate prediction of the drug-disease relationship is realized through multi-modal feature interaction, sub-graph enhancement and a dynamic sub-graph learning mechanism.

The invention provides a deep learning-based avalanche early warning model construction method and system, and the method comprises the steps: firstly obtaining multi-source environment monitoring data, including meteorological time sequence, topographic space and accumulated snow layer physical data, of a target region, carrying out the time dimension alignment of the meteorological time sequence data to generate a feature sequence, carrying out the meshing of the topographic space data to generate a feature set, and carrying out the construction of an avalanche early warning model; the method comprises the following steps: extracting parameters from accumulated snow layer physical data to generate a state vector, inputting a deep learning network model containing time sequence attention, spatial convolution and cross-modal interaction units, generating a fusion feature vector, constructing a training set based on historical avalanche event annotation data, performing dynamic weight optimization on the fusion feature vector, and generating an avalanche risk prediction model. And finally, receiving current monitoring data in real time, outputting a risk level and an early warning trigger threshold value by the avalanche risk prediction model, and generating a multi-level early warning signal when a real-time risk value exceeds the threshold value, thereby realizing accurate avalanche early warning.

The invention discloses a surgical robot resection navigation method based on multi-modal image registration. The method comprises the following steps: acquiring a preoperative multi-modal medical image of a patient and reconstructing a three-dimensional model containing a focus and a dangerous structure; acquiring an intraoperative real-time image; extracting cross-modal invariant features from the pre-operative and intra-operative images by using a first deep learning network, and establishing an initial corresponding relation; predicting an initial non-rigid deformation field by using a second deep learning network, based on the pre-operative / intra-operative image and the initial corresponding relation and in combination with physical model constraints; tracking a surgical tool in real time, and updating a non-rigid deformation field in real time by adopting an incremental calculation strategy; applying the updated deformation field to the preoperative three-dimensional model to generate a deformed model; and mapping the tool position to the deformed model, and generating a navigation instruction. According to the method, soft tissue deformation can be accurately compensated, multi-modal image information is effectively fused, navigation information is dynamically updated in real time, and the precision and safety of robot surgery are remarkably improved.

The invention relates to the technical field of power station operation and maintenance, and discloses a power generation equipment intelligent operation and maintenance method and system based on model autonomous selection, and the method comprises the steps: obtaining the multi-mode operation and maintenance data of a photovoltaic power station, and generating a multi-mode operation and maintenance data set; inputting each modal data of the multi-modal operation and maintenance data set into a corresponding module for feature extraction; inputting the extracted feature vectors into a contrast learning network for cross-modal alignment, and outputting an executable decision result by combining a retrieval enhancement generation technology with a knowledge graph; and constructing a privacy protection training framework through federated learning, inputting an executable decision result into a digital twin system for strategy verification, and generating a trained multi-modal large model for the photovoltaic power station to select a corresponding module in the trained multi-modal large model based on the feature data for real-time monitoring. According to the invention, the problems of poor accuracy and untimely reaction in the traditional operation and maintenance process of the photovoltaic power station are solved, and the operation and maintenance of the power station can be carried out timely and accurately.

The invention discloses a financial network security defense method and system based on multiple Agents and a dynamic large model. A detection Agent is deployed in an edge layer, financial network node flow data and system logs are collected in real time, time sequence features are extracted through a lightweight convolutional network, and a preliminary anomaly score is generated. And the cloud layer constructs a decision Agent, receives the feature abstract transmitted by the edge node in an encrypted manner, inputs the feature abstract into a dynamic large model for multi-modal feature fusion, and outputs defense action probability distribution. And the intelligence Agent constructs a cross-institution federated learning network. And constructing a dynamic game engine, constructing a revenue matrix based on the attack cost and the defense revenue, solving a Nash equilibrium strategy, and generating an optimal defense instruction set. And dynamically allocating detection tasks according to the threat level and the edge computing power state. According to the method, efficient acquisition and analysis are realized, the abnormal behavior recognition capability is improved, support is provided for making a defense strategy, the defense strategy is optimized, and the intelligent, automatic and efficient levels of defense are improved.

The invention discloses an image super-resolution reconstruction method based on deep learning, and belongs to the technical field of image signal processing. Comprising the following steps: S1, realizing multi-dimensional feature representation of a low-resolution input image by using a deep learning network; s2, realizing accurate alignment and lossless mapping of cross-scale features, and effectively solving the problems of feature dimension inconsistency and information distortion; s3, explicitly learning an increment mapping relation between low-resolution input and a reconstructed high-resolution image, and relieving the gradient disappearance problem of a deep learning network structure; s4, enhancing the naturalness and detail fidelity of the reconstructed image through adversarial training; s5, integrating pixel-level reconstruction loss, perception loss and structural similarity loss, and effectively balancing objective indexes and subjective visual quality of the reconstructed image; and S6, realizing joint optimization of deep learning network parameters through a unified gradient back propagation mechanism, and finally outputting a high-quality super-resolution reconstructed image.

The invention provides a tree identification method and system based on ground-air fusion three-dimensional laser point cloud, and belongs to the field of environment remote sensing data identification, and the method comprises the steps: obtaining laser radar point cloud data in a set forest region through employing an unmanned plane, and collecting the point cloud data in the set forest region through employing a laser scanning device; preprocessing the two groups of data to obtain preprocessed laser radar point clouds, and performing ground-air data seamless fusion on the preprocessed laser radar point clouds to obtain fused data; generating a canopy height model based on a watershed segmentation algorithm, and performing individual tree segmentation on the fused point cloud data by using the model; extracting a vector boundary after single tree segmentation, distributing a tree ID for each tree, and extracting a single tree point cloud according to the tree ID; the single tree point cloud is further processed to construct a multi-angle point cloud sample data set; and training a deep learning network model by using the multi-angle point cloud sample data set, and detecting the tree species to be detected by using the trained deep learning network model.

The invention discloses an amphibious unmanned aerial vehicle riverway sediment concentration real-time monitoring method based on underwater light field imaging and cross-modal fusion, and relates to the technical field of riverway sediment monitoring. Comprising the following steps: acquiring aerial multispectral data and preprocessing the aerial multispectral data; an integrated polarized light field camera is carried on an amphibious unmanned aerial vehicle to capture underwater light field data, and underwater three-dimensional light field information is reconstructed and sediment scattering characteristics are extracted through an optical model and a lightweight deep learning network; the air multispectral data and sediment scattering characteristics are fused through space-time alignment and an attention mechanism, a sediment concentration prediction model is generated, and then a spatial distribution diagram of the river sediment concentration is output; and carrying out real-time monitoring and early warning on the river sediment concentration according to the spatial distribution diagram of the river sediment concentration through edge calculation and cloud collaboration architecture. The method is combined with an advanced deep learning algorithm, high-resolution silt concentration inversion can be realized, and the method has real-time performance and adapts to a complex water body environment.

The invention discloses a short-time rainfall prediction method based on radar image and reanalysis data fusion, and the method comprises the following steps: collecting radar images and reanalysis data at continuous times, and generating input data in a unified grid format through spatial interpolation, time alignment and standardization processing; respectively extracting spatial and temporal features of the radar image and the reanalysis data by using a dual-channel encoder, and carrying out weighted fusion through a channel attention mechanism to generate a fusion feature tensor; inputting the fusion features into a ConvLSTM (Convolutional Long Short-Term Memory Neural Network), modeling a spatio-temporal evolution process of a rainfall system, and outputting a preliminary rainfall prediction image in 0-3 hours in the future; constructing a residual learning network, and performing deviation correction on the preliminary prediction result based on historical residual and observation information; when the radar image input is missing, the completeness of the input structure is maintained through the replacement feature generation module; generating a rainfall intensity image or a probability graph in 0-3 hours in the future; the method supports visual output.

The invention relates to the technical field of computer vision, in particular to a plant leaf disease severity assessment method based on position-guided focus characterization learning, which comprises the following steps: acquiring a disease leaf image; the disease leaf image is input into a hierarchical lesion fusion evaluation network model, a leaf disease severity evaluation result is obtained, the hierarchical lesion fusion evaluation network model is obtained based on training of a real training set, the real training set is a leaf image marked with the disease severity, and the leaf disease severity evaluation result is obtained; the hierarchical lesion fusion evaluation network model comprises a feature encoder with priori knowledge, a hierarchical lesion feature fusion module and an evaluation module, the feature encoder with priori knowledge is obtained by inputting an image pair into a position-guided lesion characterization learning network for training, the image pair refers to health and disease states of the same leaf, and the evaluation module refers to the position-guided lesion characterization learning network. The leaf disease state is a lesion leaf generated from a healthy leaf through a diffusion model. According to the invention, more accurate positioning and severity evaluation can be carried out on the focus area.

The invention provides a multi-modal sentiment analysis method based on unwrapping representation learning, which comprises the following steps: acquiring a multi-modal sentiment analysis data set, and extracting features of texts, audios and videos; constructing a representation learning network, extracting private features and shared features, and processing through adversarial learning optimization and time smoothness constraint; under the guidance of different language scales, the private features of the audio and the video are fused; further fusing text, audio and video features through cross-modal attention and a gating mechanism; and finally, inputting the fusion features into an emotion analysis module, and generating an emotion prediction result in combination with an emotion classification or regression task. According to the method, the consistency of the audio and video features in continuous time is considered, redundant information is effectively reduced, the time dimension consistency is ensured, deep fusion of the audio and video features is realized under text guidance, and the cross-modal fusion effect is remarkably improved.

The invention provides an air water production intelligent prediction and adjustment method and system based on meteorological big data, and relates to the technical field of air water production, and the method comprises the steps: collecting data through a meteorological sensor, constructing an enhanced feature space, fusing multi-domain knowledge through a migration cross-domain learning network, and predicting the water production through combining with a deep neural network. A control strategy is generated based on a deep reinforcement learning model, optimal control parameters are selected through multi-target Bayesian optimization and a dynamic decision algorithm, the air water production amount can be accurately predicted, intelligent adjustment is achieved, the water production efficiency is improved, energy consumption is reduced, and high environmental adaptability and robustness are achieved.

The invention discloses a medical image super-resolution reconstruction method based on multi-level attention guidance, and the method comprises the following steps: S10, constructing a deep learning network model based on a generative adversarial network architecture, which comprises a generator and a discriminator; the generator is based on an improved U-Net architecture, a hierarchical attention module and a dual-path feature processing module are configured in an encoder and a decoder of the generator, the hierarchical attention module adopts different attention strategies according to network levels to consider structure and texture, and the dual-path feature processing module separates and processes low-frequency and high-frequency information; the generator further comprises a multi-level feature fusion module for integrating the multi-scale features of the decoder, and an attention guide up-sampling module for final enhancement and dimension raising. The discriminator adopts a spectrum normalization U-Net architecture and uses multi-scale features for matching; s20, training the network model by adopting a composite loss function comprising pixels, adversarial, perception and total variation loss; and S30, inputting the low-resolution image into the trained model, and outputting a high-resolution image. According to the method, through deep fusion of multi-level attention and multi-scale feature processing, the image restoration quality can be remarkably improved, the texture detail definition can be enhanced, the anatomical structure accuracy can be ensured, and the noise robustness can be improved.

The invention discloses a visual servo double-arm robot migration simulation learning method from simulation to reality, which comprises the following steps: S110, constructing a virtual simulation scene according to a real task scene, and establishing a relation between the virtual simulation scene and the real task scene; s120, the collected teaching mechanical arm and executing mechanical arm operation data are replayed and optimized in the virtual simulation scene, so that a data set used for final training is constructed; s130, a deep imitation learning network is designed and achieved, input of the deep imitation learning network comprises visual data obtained in the real task scene and state data of all joint motors of the teaching mechanical arm, and output of the deep imitation learning network is predicted states of all joint motors of the execution mechanical arm at the next moment; and S140, migrating the fully trained and converged deep imitation learning network from a virtual simulation scene to a double-arm robot in a real task scene. According to the invention, the exploration efficiency and generalization ability of the two-arm robot learning network are improved.

The invention provides a zero-sample dynamic MRI reconstruction method based on global and local diffusion models, and the method comprises the following steps: generating an image data set based on original two-dimensional dynamic full-sampling heart magnetic resonance multi-coil data; designing a global and local time interlaced sampling strategy to construct two types of training data sets for generating a network model by adopting global and local cascade learning networks of a two-stage learning mechanism; performing prediction, correction, low-rank constraint and data consistency on data alternation by using an iterative solution algorithm for rapidly extracting prior information of a network model; and applying an iterative solution algorithm to the trained network model, testing, outputting a dynamic magnetic resonance reconstruction image, and evaluating the image quality. According to the method, the complete coding data is constructed through different time interlaced sampling strategies, the image is quickly and accurately reconstructed under the under-sampling condition in combination with the image prior information of the network model, and the quality and efficiency of the reconstructed image are remarkably improved.

The invention relates to the technical field of autonomous navigation and robot environment perception, and discloses a multi-source sensor fused adaptive navigation system, which comprises a multi-source sensor space-time synchronization module, a quantum particle filtering positioning estimation module, a space-time element learning controller module, a cross-modal quantum fusion module and an adaptive navigation control module. Multi-source data space-time alignment is realized through Lie group SE (3) calibration and dynamic time warping; the positioning robustness of particle filtering is improved based on quantum state coding and annealing optimization; dynamically distributing a fusion weight and injecting a physical constraint by utilizing a meta-learning network; feature level fusion of laser radar, vision and inertial data is realized by means of a quantum entanglement mechanism; and constructing closed-loop adaptive navigation by combining model predictive control and quantum purity trigger feedback. According to the method, the navigation reliability problem caused by misalignment of multi-modal sensor data fusion, divergence of state estimation and insufficient cross-modal relevance in a dynamic environment is solved.

The invention provides an equipment corrosion thinning life prediction method and system, and relates to the technical field of life prediction, and the method comprises the steps: deploying edge computing nodes at a plurality of equipment sites to collect corrosion data, carrying out the encryption through employing a differential privacy algorithm, and constructing a federated learning network for model training; and the central server extracts prediction experience through knowledge distillation to form knowledge priori, the edge nodes predict the corrosion rate and the residual life based on real-time data, and the server generates a comprehensive evaluation report and continuously optimizes the model. According to the method, safe sharing and collaborative prediction of the equipment corrosion data are realized, the prediction precision is improved, and the data privacy is protected.

The invention relates to the technical field of power system monitoring and fault diagnosis, and discloses a partial discharge detection network system based on an adaptive attention mechanism. Comprising a data acquisition module, a signal preprocessing and time-frequency image generation module, a deep learning network of an adaptive attention mechanism, a fault classification and diagnosis module and a real-time monitoring and alarm module, the data acquisition module acquires a multi-mode partial discharge signal in real time, and preprocessing is performed through the signal preprocessing and time-frequency image generation module. The deep learning network of the adaptive attention mechanism carries out feature extraction and classification on the time-frequency image, finally, the fault classification and diagnosis module outputs a diagnosis result, and the real-time monitoring and alarm module carries out real-time monitoring and automatic alarm according to the diagnosis result. According to the system, multi-mode partial discharge signals are collected in real time, feature extraction and classification are carried out through a deep learning network, the attention degree of different features can be dynamically adjusted through a self-adaptive attention mechanism, and the accuracy and robustness of detection are improved.

The invention discloses a federated learning-driven cross-domain supply chain elastic inventory optimization system and a method thereof, and aims at realizing inventory data collaboration among same-level enterprises or regional nodes through transverse federated learning and ensuring data security by adopting a self-adaptive differential privacy protection mechanism. The method comprises the steps of constructing a transverse federated learning network, locally performing data preprocessing, adding differential privacy noise, iteratively training a global model based on federated deep reinforcement learning, generating a transverse inventory allocation and replenishment decision, and performing model adaptive adjustment in real time based on key performance indicators. According to the method, multi-target balance is considered, inventory configuration is dynamically optimized through a multi-target reward function, the inventory turnover rate is remarkably increased, the inventory holding cost is reduced, the service level is improved, and the method is suitable for various scenes such as retail chain, manufacturing industry distributed storage and cross-regional logistics distribution; and global optimal inventory configuration is realized on the premise of ensuring data privacy.

A method discloses receiving a plurality of input images, receiving text prompts, generating a visual matrix utilizing the images and an image encoder, generating a text matrix utilizing a text encoder, multiplying the text matrix and the visual matrix to generate an image-text similarity matrix that assigns a numerical value indicating similarities between each of encoded visual descriptors and each of the encoded images, wherein similarities are indicated by entries of the image-text similarity matrix having numerical values that determine a loss function associated with the image-text similarity matrix, identify a gradient of the loss function with respect to parameters associated with the image encoder and parameters associated with the text encoder, utilizing the gradient, update parameters associated with the image encoder or the text encoder, and outputting final updated parameters associated with either the text encoder or image encoder of the machine learning network.