1302 results about "Attention network" patented technology

The invention discloses a software multi-agent collaboration method and system based on a large language model, and the method comprises the steps: receiving natural language task description submitted by a user at the same time, carrying out the semantic understanding and intention recognition through a pre-trained large language model center, and generating a structured task element set; based on the structured task element set, the large language model center generates a task dependency graph through multiple rounds of reasoning, and the task dependency graph comprises a plurality of atomic subtasks, logic relations among the tasks and data flow constraints; according to a topological structure and resource demand characteristics of a task dependency graph, a double-layer graph attention network is adopted to dynamically match a professional agent with specific domain capability, and a distributed collaborative network is formed. Through the dynamic graph network scheduling and cross-domain semantic alignment mechanism, the problems that the multi-agent dynamic collaborative adaptation capability is insufficient and cross-domain semantic fusion is difficult are solved.

The invention belongs to the technical field of remote sensing super-resolution images, and particularly relates to a remote sensing image super-resolution system and method based on an adaptive Mamba-attention network. Comprising a feature extraction module used for carrying out shallow feature extraction on an input low-resolution image to obtain shallow features; the multiple cascaded adaptive state space blocks are used for processing the shallow layer features to obtain reconstruction features; and the reconstruction module maps the reconstruction features to a target resolution space through sub-pixel rearrangement operation to obtain a high-resolution remote sensing image. High-frequency details and a low-frequency structure are cooperatively processed in a feature space by using the remote sensing frequency sensing modulation module, and high-resolution output is generated by combining sub-pixel rearrangement up-sampling, so that high-quality reconstruction of a complex remote sensing scene is realized.

The invention discloses a cloud side-end cooperative task scheduling and efficiency optimization method and system for heterogeneous patrol resources, and relates to the technical field of intelligent scheduling and resource optimization. According to the method, accurate perception of a resource state is realized by constructing a digital twinborn and federated learning mechanism, resource contention conflicts are solved by adopting a space-time diagram attention network and multi-agent reinforcement learning, and multi-target optimization and trusted execution are realized in combination with a quantum genetic algorithm and a block chain smart contract. Finally, the stability of the system is verified through Lyapunov optimization, a complete scheduling system from resource perception and conflict resolution to steady state maintenance is formed, and the task scheduling efficiency and the system stability in the heterogeneous resource environment are remarkably improved.

The invention provides a self-adaptive bearing fault diagnosis method based on multi-base wavelet fusion. The objective of the invention is to solve the problems of noise reduction, insufficient feature extraction and low diagnosis precision under noise conditions. A Kaisixi University bearing public data set is used as original data, and Gaussian noise with different SNRs is superposed to simulate various noise intensities. And uniformly carrying out length alignment, down-sampling, equal-length segmentation, division and normalization preprocessing. Then, wavelet bases such as sym4, db4, coif5 and the like are adopted for parallel multi-scale decomposition and reconstruction; and adaptively determining the number of decomposition layers and a threshold strategy according to the noise level, and generating a de-noising branch. And performing weighted fusion on the denoising results of the branches, and performing iterative denoising on the residual error. Signals subjected to noise reduction processing are sent to a double-branch convolution-cycle-attention network, a convolution layer extracts features, an LSTM and a self-attention module capture time sequence changes, and accurate recognition of various bearing faults is achieved. The training adopts a segmented attenuation learning rate and an early stop strategy, and the robustness and generalization ability of different SNR working conditions are improved.

The embodiment of the invention provides a dynamic flexible workshop scheduling method and related equipment, and belongs to the technical field of industrial intelligent manufacturing and production scheduling. The method comprises the steps of obtaining current state information of a workshop in response to a scheduling event; inputting the state information into a pre-trained scheduling decision model for processing; the model outputs state feature embedding through a two-stage feature extraction network: in the first stage, feature extraction is performed on a heterogeneous disjunction graph by using a graph attention network, and in the second stage, expert output is dynamically fused through a hybrid expert model; and finally, outputting and executing a process-machine pairing decision by the actor network. Wherein the model is trained by adopting a near-end strategy optimization algorithm based on multiple commentators; and a meta-learning framework is integrated during training, so that the model obtains strong generalization ability. According to the method, the defects of an existing scheduling method in the aspects of state characterization, multi-target tradeoff and environmental adaptability are effectively overcome, and the efficiency, quality and robustness of dynamic flexible workshop scheduling are remarkably improved.

The invention relates to the technical field of natural language processing, in particular to an NLP-based multi-source data fusion system and method.The NLP-based multi-source data fusion system comprises a data collecting and processing unit, an NLP processing unit, a data fusion unit and an output unit.The data collecting and processing unit collects multi-source heterogeneous data and executes standardization processing to obtain preprocessed data; the NLP processing unit extracts entity, relation and emotion features through deep semantic analysis, a context-aware entity relation graph is constructed by adopting a bidirectional attention mechanism model and a graph attention network, and the data fusion unit realizes cross-source entity ambiguity resolution through an iterative graph neural network based on a graph topological structure. The emotion confidence weight is dynamically distributed to generate a fusion vector, a rule feedback reconstruction map is extracted, and the output unit converts the fusion vector into a target format for output, so that the problem of semantic conflict of multi-source data is solved, and semantic coherence and availability of fusion data are improved.

The invention relates to the technical field of computer vision, and particularly discloses a CNN and Transform fused self-supervised monocular depth estimation system and method. According to the system, local representation is enhanced through a multi-scale feature fusion mechanism, a cross-regional attention network is constructed to realize global context association, and fine reduction of a fine structure of a complex scene is realized. Firstly, based on DCB, multilayer expansion convolution is adopted to expand a receptive field, multi-scale pixel features are fused, and local details of a key area are enhanced; secondly, capturing fine-grained local information of the image by using parallel local convolution of ELGF, and acquiring long-distance dependency by using a self-attention mechanism, thereby realizing collaborative modeling of local information and global dependency, and remarkably enhancing feature expression ability; and finally, estimating a relative pose between adjacent images through a pre-trained ResNet18-based lightweight encoder, and constructing reprojection loss to optimize depth prediction. Experiments show that the model constructed by the method provided by the invention reaches 0.102 and 4.430 in AbsRel and RMSE indexes respectively, and is obviously superior to the existing mainstream method.

The invention belongs to the field of artificial intelligence and electroencephalogram emotion recognition, and provides a double-branch electroencephalogram emotion recognition method and system based on brain region topology and time-space, and the method comprises the steps: preprocessing a to-be-recognized electroencephalogram signal to obtain a plurality of electroencephalogram fragments, and extracting a difference entropy sequence of each electroencephalogram fragment and a Spearman correlation coefficient matrix between channels; based on the Spearman correlation coefficient matrix, utilizing a bridging dynamic graph attention network module to extract topological features of a brain region; processing the differential entropy sequence by using a multi-scale space-time mixed attention module to obtain multi-scale space-time features; carrying out residual mutual cross attention fusion on the topological features of the brain region and the multi-scale spatial-temporal features to obtain fusion features; and performing classification based on the fusion features, and determining an emotion recognition result corresponding to the electroencephalogram signal. According to the method, the accuracy and robustness of emotion recognition are improved by utilizing the spatial topology characteristics and the multi-topology time dynamic characteristics of the electroencephalogram signals, and the defects of modeling spatial dependence and time dynamic are overcome.

The invention discloses an attention perception path reasoning method for a knowledge graph, and the method comprises the steps: firstly, generating an entity representation containing local semantics through employing a graph attention network coding entity and an adjacency relation, and synchronously obtaining an attention weight representing the association intensity between entities; secondly, innovatively providing a target-guided biased random walk path sampling strategy, and adaptively exploring a high-quality multi-hop semantic path related to a target task by taking the attention weight as a bias; then, information aggregation is carried out on the sampled semantic paths through a path encoder, and global path representation is obtained; and finally, carrying out deep fusion on the local entity representation and the global path representation, and jointly inputting the local entity representation and the global path representation into a prediction layer to carry out knowledge graph link prediction. According to the method, local structure perception and global path reasoning are cooperatively optimized through an attention mechanism, so that the link prediction precision and interpretability of the knowledge graph are remarkably improved.

The invention relates to the technical field of computers, discloses a product full-process quality monitoring method and system based on big data, and aims to solve the problem of insufficient quality management refinement caused by data islands, monitoring lag, tracing difficulty and single analysis dimension in the prior art. The method comprises the following steps: constructing a full-process data acquisition system covering design, materials, production, logistics and after-sale, integrating multi-source heterogeneous data, and performing cleaning and time alignment; using an LSTM and GRU mixed model to extract time sequence features of the process parameters, and generating a high-dimensional process feature vector; and constructing a material-process-quality association knowledge graph in combination with the graph attention network, and representing the internal association of the quality influence factors of each link. According to the scheme, full-process data fusion, active quality prediction, accurate traceability and intelligent optimization are realized, and the real-time performance, the accuracy and the self-adaptive capability of quality control are remarkably improved.

The invention provides a circuit board electroplating process quality monitoring and early warning method, which comprises the following steps: carrying out real-time acquisition and standardization processing on process parameters (such as current density, bath solution temperature and pH value) of procedure nodes of acid pickling, activation, electroplating and the like by utilizing a distributed data acquisition system, and fusing historical causal relationships based on a knowledge graph structure to obtain a circuit board electroplating process quality monitoring and early warning result; according to the method, causal strength dynamic evaluation is realized by combining improved Granger test and a dynamic Pearson coefficient, long-range dependence between multi-hop graph attention network modeling parameters is adopted, an optimal abnormal attribution path is further screened through path credibility attenuation and a trend semantic encoder, and finally, an optimization suggestion template is automatically matched to output a process adjustment scheme, so that the accuracy of the abnormal attribution path is improved. According to the invention, the causal reasoning accuracy and the automatic optimization capability of the electroplating process anomaly analysis are improved, and the operation reliability and the intelligent level are improved.

The invention provides a storage resource optimization method and system based on a time sequence dependence hypergraph neural network, and belongs to the field of artificial intelligence computing. Based on medical health big data and computing resources, constructing and fusing data and computing resource dependency matrixes to obtain a static dependency relationship matrix; the method comprises the following steps: collecting a computing resource multi-source operation log, generating dynamic characteristics of each moment according to a fixed interval, introducing time sequence position coding and self-attention mechanism weighting in a sliding time window to obtain attention optimization characteristics, and generating a dynamic dependency weight matrix by combining a modeling historical hidden state and the dynamic characteristics; the static and dynamic dependency weight matrixes are fused to obtain a comprehensive dependency matrix, attention optimization features are used as nodes, hyperedges are constructed in combination with the comprehensive dependency matrix, and hypergraph association and other matrixes are generated; and inputting the matrix into a graph neural network, learning node representation in combination with a time sequence attention network, classifying nodes and mapping the nodes into scheduling actions, and realizing self-adaptive allocation of storage resources in combination with target function optimization of medical scene constraints.

The invention discloses an automatic UI self-healing test method and system based on multi-agent AI, and belongs to the technical field of automatic UI tests.The automatic UI self-healing test method comprises the steps that an original test script is semantized through a test reasoning interpretation class, and an expected intention is output; when the test fails, synchronously collecting multi-source data and extracting multi-modal features, constructing a fault propagation graph according to an expected intention and the multi-modal features, and performing fault propagation analysis by using a graph attention network to obtain a fault diagnosis result; according to a fault diagnosis result, a deep Q network, a greedy strategy and a Byzantine fault-tolerant algorithm are adopted to negotiate and decide a repair strategy through a distributed agent cooperation mechanism, and feedback is obtained; optimizing and dynamically adjusting the repair strategy and the test reasoning interpretation class through a meta-learning algorithm based on feedback; a test intention is deeply understood by constructing a test reasoning interpretation class, and fault diagnosis and automatic repair are cooperatively completed by adopting a distributed intelligent agent, so that the semantic understanding capability, the diagnosis accuracy and the repair success rate are improved.

The invention relates to the technical field of data processing and mode recognition, in particular to a bearing fault diagnosis method based on multi-scale feature fusion, which comprises the following steps: fusing multi-source data such as vibration, acoustic emission and rotating speed, performing angle domain resampling by using rotating speed data, generating a two-dimensional order spectrogram, and stacking to construct a three-dimensional working condition information tensor; a master-slave modulation heterogeneous neural network is adopted, high-dimensional spatial-temporal features are extracted through a main branch three-dimensional convolutional network, time sequence details are extracted from an original sequence through an auxiliary branch one-dimensional convolutional network, affine transformation parameters are generated, and dynamic modulation is achieved on the high-dimensional features; and the output state vector is mapped to a fault evolution knowledge graph, probability prediction is carried out through a graph attention network and by introducing a Monte Carlo discarding mechanism, a probability mean value is calculated as a fault classification result, and the diagnosis confidence is quantified by a probability variance. According to the invention, through multi-scale feature fusion and dynamic modulation, the problem of insufficient feature discrimination caused by scale mismatch under variable working conditions is solved.

The invention relates to the technical field of computer vision, in particular to a space-time consistency data generation method for visual target tracking. The method comprises the following steps: firstly, training a path generator on a target tracking training set, learning a motion law of a target in a time sequence by using optical flow estimation and conditional variation coding technologies, and generating a target motion track conforming to physical constraints; and then, based on the generated target trajectory, introducing a space-time consistency attention mechanism to guide a text-video generation model, and under the condition of keeping basic model parameter freezing, constraining the position, scale and continuity of a target in a generation frame through an attention network, thereby synthesizing a video frame sequence with real motion features. According to the method, target tracking video data with real motion characteristics and high time sequence consistency is generated, and the robustness of the model to complex motion, illumination change and shielding conditions can be improved in different scenes.

The invention provides an electroencephalogram-myoelectricity collaborative limb movement intention decoding method and system based on a symmetric cross-modal attention network, and belongs to the technical field of neural rehabilitation engineering and movement intention decoding. Comprising the following steps: acquiring an electroencephalogram signal and an electromyographic signal to be decoded; inputting the electroencephalogram signals into an electroencephalogram channel specificity feature extraction network, and extracting multi-scale space-time oscillation features; the electromyographic signals are input into an electromyographic channel specificity feature extraction network, and dynamic time sequence mode features are extracted; inputting the electroencephalogram features and the myoelectricity features into a symmetric cross-modal attention module, carrying out bidirectional feature interaction and calibration, and generating fusion enhancement features; the symmetric cross-modal attention module realizes mutual enhancement and alignment between the electroencephalogram signals and the electromyographic signals by taking own features as queries and taking another modal feature as a key and a value through a cross attention mechanism; and inputting the fusion enhancement feature into a classifier, and decoding to obtain a corresponding motion intention category.

The invention relates to the technical field of knowledge maps, and discloses a spatio-temporal knowledge map establishment method, device and equipment and a storage medium, multi-source spatio-temporal data is preprocessed, the preprocessed multi-source spatio-temporal data is input, pre-trained and injected into a large language model with spatio-temporal perception ability, and a spatio-temporal knowledge map is established. Outputting an entity with space-time confidence, a relation triple and a space-time trajectory fragment thereof; performing knowledge fusion on an output result by adopting a space-time multi-mode graph attention network to form a space-time knowledge graph; when a new data stream enters, real-time updating of the space-time knowledge graph and entity space-time trajectory recording are carried out; on the spatio-temporal knowledge graph, link prediction and event deduction are carried out by utilizing a spatio-temporal graph neural network, reinforcement learning decision optimization is combined, a spatio-temporal reasoning result and a decision scheme are output, and services are provided for an application system through an API interface; the dynamic spatio-temporal information processing capability is improved, and the knowledge graph updating efficiency is improved.

The invention relates to the technical field of unmanned aerial vehicles and unmanned vehicles, in particular to an air-ground unmanned cluster conflict resolution method based on multi-agent reinforcement learning, which comprises the following steps of: establishing an air-ground unmanned cluster navigation environment, determining obstacles and a plurality of air-ground subgroups, and determining that routes from each air-ground subgroup to a target navigation point have conflict positions; determining a state space and an action space of an air space subgroup and a reward and punishment function of an air space subgroup action based on the air space unmanned cluster navigation environment; establishing strategy networks including a local Q network, a local strategy network, a hybrid network and an attention network; determining a network loss function; performing reinforcement learning training on the strategy network to obtain a trained strategy network; controlling the air-ground subgroup to navigate by a decision action output by the trained strategy network; according to the invention, the conflict of multiple groups of clusters can be resolved.

The invention relates to the technical field of online approval management, and discloses an online employee approval management system based on an edge calculation model. An employee terminal data acquisition layer of the system captures a multi-modal examination and approval data flow based on a priority dynamic focusing mechanism, generates an examination and approval behavior characteristic thermodynamic diagram through an edge characteristic pyramid network, and divides an abnormal region; the examination and approval knowledge graph construction layer loads a domain rule base, extracts an entity relationship through semantic dependency analysis, and constructs an examination and approval rule path graph with a weight; the map path cross validation layer synchronizes an edge node clock, and generates approval risk confidence through multi-head map attention network fusion data; the dynamic strategy optimization layer converts the risk confidence into an executable strategy and issues the executable strategy to an edge execution engine; and the approval efficiency feedback layer monitors the flow state and generates an efficiency index to dynamically optimize a data acquisition strategy. According to the system, precision, high efficiency and dynamic optimization of approval management are realized.

The invention provides an artificial intelligence semantic processing system and method oriented to digital media creation, and relates to the technical field of artificial intelligence semantic process.The artificial intelligence semantic processing method comprises the steps that predicate argument relation pairs of language texts are extracted, object space relation pairs of sketch images are extracted at the same time, and a basic semantic unit set is constructed; the integrity and accuracy of cross-modal semantic understanding are ensured, further, semantic units are clustered by using a dynamic routing algorithm, a semantic concept cluster with a clear importance weight is generated, deep mining and structured representation of creation intentions are realized, and the creation intentions are quickly and accurately understood. An initial semantic relation graph is constructed, a graph attention network is used for dynamic reweighting, finally, an enhanced dynamic semantic graph is generated, complex association and a hierarchical structure between semantic concepts are effectively captured, finally, hierarchical analysis is carried out on the semantic graph, and a structured semantic blueprint is output, so that the dynamic semantic graph is obtained. And a reliable semantic processing technology is provided for creation of high-quality digital media contents.

The invention relates to the technical field of artificial intelligence, and discloses a Transformer training method for paint surface AI defect identification, which comprises the following steps: synchronously acquiring RGB images and three-dimensional point cloud data through a six-axis mechanical arm to form a multi-modal data set; generating synthetic defect data by using fluid dynamics and a ray tracing model to expand a sample set; realizing feature alignment by adopting cross-modal contrast learning; aggregating defect structure capsules through a dynamic routing attention network; performing knowledge retrieval and fusion in combination with the prototype memory matrix; and finally, the sorting mechanical arm is driven to execute sorting or rechecking operation based on the multi-task output and the uncertainty score. According to the method, the problems of insufficient structural representation of complex defects, weak generalization ability of rare defects and low reliability of sorting decision are solved, and high-precision and high-reliability automatic defect detection and sorting are realized.

The invention discloses a multi-modal data drawing logical relationship analysis method, electronic equipment and a medium, and the method comprises the steps: generating a node set based on drawing image data and text data; generating a cross-modal hyperedge set based on the spatial proximity relationship, the visual feature similarity and the semantic correlation between the node sets; generating a hypergraph embedding input representation based on the node set and the cross-modal hyperedge set; the hypergraph is embedded into the input representation input improved hypergraph self-attention network model, and a hyperedge logic relation type and a corresponding hyperedge confidence coefficient are generated; generating a graph structure result based on the hyperedge logic relationship type and the node set, wherein the graph structure result meets the structure legality requirement; and performing hyper-parameter automatic adjustment and convergence control on the atlas structure result based on hyper-edge confidence, and generating an optimal atlas analysis model and a structured output result. According to the method, the reliability and the quality of analysis of component nodes, logic edge relationships and semantic structures in the drawing are improved.

The invention discloses a disaster multi-source data real-time fusion early warning system based on deep learning, which belongs to the technical field of disaster early warning and comprises a multi-source data space-time alignment module, a cross-modal semantic fusion module, a disaster evolution dynamic prediction module and a minute-level early warning generation module. According to the method, accurate alignment of multi-source data is realized through an adaptive space-time resampling algorithm, cross-modal deep semantic fusion is realized through a cross attention mechanism, a disaster evolution trend is predicted through a space-time diagram attention network, minute-level early warning response is realized through an incremental calculation strategy, and disaster early warning is realized. According to the method, precise space-time alignment of the multi-source heterogeneous data is realized through adaptive space-time resampling and timestamp synchronous correction, and the space-time alignment precision is improved.

The invention relates to the technical field of airspace control, and discloses an AI-based low-altitude airspace information center collaborative control method, which comprises six core steps of multi-source sensing fusion, abnormal behavior discrimination, dynamic risk assessment, environmental constraint re-planning, communication collaborative optimization and collaborative execution decision-making to construct a closed-loop intelligent decision-making mechanism. Microsecond-level time synchronization is realized through a Beidou or GPS satellite time service system, and the target identification precision is improved in combination with a weighted evidence fusion theory; an R-tree and space-time bounding box dual rule engine is combined to accurately detect abnormal flight; the collision risk between the aircrafts is quantified by using a three-layer stacked multi-head map attention network, and trajectory re-planning under environmental constraints is realized through a multi-dimensional environment voxel model; a ground and satellite hybrid networking architecture is adopted for communication collaborative optimization, so that terrain shielding and electromagnetic interference are avoided in management and control communication, global coverage and accurate adjustment are realized, and the accuracy, real-time performance and safety of management and control are ensured.

The invention relates to the technical field of forestry management, in particular to a forest and grass resource dynamic monitoring system which comprises the steps that a graph attention network is adopted to conduct high-low weight recognition processing on the structure difference value between node pairs, feature vectors are established through the canopy height difference and canopy density difference between adjacent nodes, and the canopy height difference and the canopy density difference between adjacent nodes are obtained; an edge weight scoring system is constructed in the form of segmented statistics and proportion weighting, node edge pairs with high influence relation strength are dynamically screened, ordered aggregation of spatial communication strength is realized, a graph neural network is introduced in a parameter fusion stage to construct a node parameter representation mechanism, and the spatial communication strength is improved. The tree species proportion, the grade of diameter at breast height and the community vertical structure in a regional sample plot are used as input features, unified mapping of node features of each region is completed in multiple rounds of iteration, cross-regional difference analysis is executed on model output in combination with biomass change frequency, parameter items with the difference proportion lower than a set threshold value are replaced with unified expression, and the model output is obtained. And parameter synchronization is realized to realize space nesting and feature integration.

The invention discloses a multi-modal fine-grained semantic alignment method and device based on a graph neural network, and relates to the field of multi-modal deep learning. Firstly, deep feature extraction is performed on input multi-modal original data, and then word-level text features and local image features are constructed into a cross-modal graph structure. And performing weighted aggregation on node neighborhood information of the cross-modal graph structure through the graph attention network. And finally, carrying out weighted fusion on the text alignment features and the image alignment features. According to the cross-modal feature fusion method, the word-level text features and the local image features are uniformly abstracted into the graph structure nodes for refined alignment, a more accurate cross-modal semantic corresponding relation can be captured, the heterogeneity problem in expression modes and semantic structures is effectively relieved, and the accuracy and reliability of cross-modal feature fusion are improved. The graph attention network can adaptively adjust the weight distribution of information propagation, highlights the effect of key features in the alignment process, and ensures that the model makes full use of important semantic relationships.

The invention provides a quadruped robot self-adaptive following method and device based on multi-modal space-time fusion, a quadruped robot and a storage medium, and the method comprises the steps: obtaining sensing data in multiple modals based on multiple sensing devices deployed on the quadruped robot, performing time and space alignment on the sensing data in various modes to obtain aligned sensing data; performing feature extraction on the aligned sensing data in the multiple modes to obtain feature data in the multiple modes, and performing attention processing on the feature data in the multiple modes by using a pre-trained multi-head attention network to obtain association degrees among the multiple modes; based on the correlation degrees corresponding to the multiple modes respectively, motion track planning of the automatic following target object is carried out; and under the condition that the target object is lost in the sensing data in any mode, executing motion compensation so as to obtain the target object again in the mode where the target object is lost.

The invention discloses an anti-fact multi-mode dialogue emotion causal reasoning method based on a double-branch hypergraph. The method comprises the following steps: respectively extracting sentence level feature vectors of three modes of text, voice and vision from input multi-mode dialogue data; carrying out modeling on a high-order relationship in the modals and between the modals by utilizing a hypergraph structure, and constructing a dialogue hypergraph containing multi-modal nodes and emotion nodes; introducing a hypergraph attention network on the hypergraph, learning contribution weight of each modal node to a target emotion node, and selecting a candidate reason node set; the candidate reason nodes are intervened, an anti-fact branch is constructed, a fact situation and final node feature representation under the anti-fact situation are calculated, and a causal effect vector is obtained; and designing a joint optimization objective function, and carrying out joint training on emotion recognition loss and causal consistency loss to realize synchronous prediction of emotion categories and emotion reasons. According to the method, a high-order semantic relationship can be effectively modeled in a multi-modal dialogue scene, and a key reason for emotion formation is reasoned.

The invention relates to the technical field of electric digital data processing, and discloses an intelligent data processing method for clinical research. The method comprises the following steps: constructing a basic clinical knowledge graph fused with an international medical ontology; the method comprises the following steps: performing deep learning driven entity recognition and knowledge graph linking on original data from heterogeneous sources such as an electronic medical record and an inspection system; executing cross-source entity alignment and knowledge fusion based on the graph attention network; performing automatic data quality verification and restoration according to clinical logic rules embedded in the atlas; and finally, extracting and generating a standardized analysis ready data set from the enhanced atlas according to research requirements. According to the technical scheme, high-quality, automatic and semantic integration and flexible delivery of clinical research data are realized, and the data processing efficiency and research reliability are improved.

A non-signalized intersection automatic driving vehicle safety decision-making method based on deep reinforcement learning comprises the steps that firstly, time sequence states of a vehicle and other adjacent vehicles are obtained through vehicle and environment interaction, the time sequence states are input into a time-space social attention network, and time social attention captures the time sequence dependency relation of the vehicles through a one-dimensional convolution and self-attention mechanism; the space social attention adopts a full-connection network and a self-attention mechanism to model a space interaction relationship between vehicles, after feature alignment and dimension fusion are performed on the two vehicles, space-time attention features are generated, and a driving strategy is output through a TD3 algorithm; a self-adaptive strategy correction mechanism is introduced, the risk margin is dynamically adjusted by taking real-time collision time as a constraint, and a danger strategy is corrected in combination with a constraint prediction model; meanwhile, the risk is evaluated through the position relation between each HVs at the intersection and the potential collision area, and the TD3 is guided to learn a safer behavior mode through the dynamic association of the risk and the reward, so that the safe and stable control of the automatic driving vehicle at the non-signalized intersection is realized.