19 results about "Automatic inference" patented technology

In one example, a method performed by a processing system including at least one processor includes automatically inferring metadata about a structured query language database, constructing, based on the metadata, a prompt for generating a new structured query language query to be executed against the structured query language database, and providing the prompt to a language model for use in generating the new structured query language query.

The invention belongs to the technical field of artificial intelligence, and particularly relates to a GUI-Agent trajectory data generation method and system based on multi-agent collaborative reasoning. The method comprises the following steps: S1, capturing and analyzing a page state and a structure, and constructing an initial graph; s2, analyzing semantics of interactive elements based on VLM and generating an initial action instruction set; s3, executing the instruction, recording state change, automatically deducing a target intention, forming an initial track, and storing the initial track in a track pool; s4, for the current track, predicting and executing the next action according to the target and the page state of the current track, and updating the track; s5, unexplored elements are recognized, new actions and intentions are generated in combination with the VLM, and a new track is created and added into a track pool; s6, evaluating whether the action result achieves a target or not, and deciding to terminate, adjust or continue to explore according to the target; and S7, circularly executing the steps S4 to S6, performing parallel iteration processing on the trajectory pool until a termination condition is met, and outputting a trajectory data set.

The invention discloses a footprint image human body attribute automatic inference method based on a multi-modal large model, and relates to the technical field of mode recognition. The inference method comprises the following steps: constructing a CLIP-based footprint image and text analysis multi-modal large model: the model adopts a double-flow network structure, a graphic encoder processes image features, a text encoder processes text features, and cross-modal feature fusion is realized through a feature fusion layer; and constructing a Grad-CAM interpretable visualization model, wherein Grad-CAM is a gradient-based feature visualization technology in deep learning. The gradient of the feature map is output through the calculation model, an attention map is generated, and key areas concerned by the convolutional neural network model during decision making are displayed; the correlation between the footprint image and the human body attribute is deeply discussed, a new multi-modal feature fusion method and model architecture are provided, and the algorithm, the model and the application method thereof formed in the research process also promote the industrial application of the technology.

The application discloses a method for automatically inferring semantics of a data cleaning script, comprising loading an original table to the data cleaning script and obtaining an input table and an output table; constructing a table change space, obtaining a table change characteristic set based on comparison between the input / output table and the table change space; taking data conversion operation types that all satisfy inevitable occurrence, impossible occurrence and possible occurrence as candidate data conversion operation types; converting and splicing multiple descriptive texts corresponding to function names, function parameters and table change characteristics to obtain a semantic vector, inputting the semantic vector into a twin convolutional neural network model to obtain candidate data conversion operation type ranking; constructing a parameter slot of each data conversion operation type, inferring the parameter slot of the candidate data conversion operation type by using data cleaning code, table content and table change information of a corresponding row, and obtaining the type and parameters of the data conversion operation based on the inference result. The method can simply and conveniently infer the type and parameters of the data conversion operation in the cleaning code.

We provide the system. [Solution] A means of acquiring a large dataset including location data from multiple external data sources, A means of performing format standardization and data cleaning on acquired data, A means for automatically inferring potential causal relationships from the dataset using generative artificial intelligence, A means for generating an analysis report based on inferred causal relationships, A means of delivering and displaying the generated analysis report on a terminal, A system that includes this.

The invention relates to the technical field of databases and artificial intelligence, and discloses a data lake table association relationship automatic discovery method and system, and the method comprises the steps: generating a table description and a column description of a new table through automatic reasoning based on a knowledge base and a language model; calculating a semantic similarity and a value overlapping distance between the query column and a target column of any table in the data lake; screening out a candidate column associated with the query column based on the semantic similarity and the value overlapping distance, and considering that an association relationship exists between the candidate column and the query column; and constructing a metadata graph based on the table description, the column description, the affiliation relationship between the table and the column and the discovered association relationship between the columns, and performing multi-hop reasoning by calculating a pass closure to derive the potential association relationship between the columns. According to the method, through knowledge-enhanced metadata construction, fusion of semantic features and value overlapping features, and a mapping metadata organization and graph reasoning technology, automatic inference of the connection relation between tables in a data lake is achieved.

The application relates to the technical field of relationship graph data clustering, and provides a relationship graph data clustering method and system based on a non-parametric Bayesian method, which comprises the following steps: obtaining relationship graph data; initializing model parameters according to a probabilistic graph model; using multiple sampling algorithms to iteratively sample samples of the model parameters based on the relationship graph data; selecting a cluster division sample according to a maximum likelihood function mode based on the model parameters obtained through sampling, so as to obtain a clustering result; and the prior probability of the cluster division sample is a non-parametric Bayesian prior. The number of clusters can be automatically inferred in the parameter inference process, and manual setting is not needed.

The invention provides an advanced instruction understanding and long-range navigation method based on a visual language model, and the method comprises the steps: carrying out the semantic object reasoning and spatial position prediction based on an advanced instruction and a global three-dimensional scene with annotations through a visual language model, and obtaining a target region; determining a target waypoint in the local boundary of the target area through a random sampling mode; and based on the panoramic view of the target waypoint, through the intelligent model with the body, carrying out detection and positioning to obtain the position information of the target object. Compared with the large difference between a navigation task and a real world demand in the prior art, the method comprises the following steps: performing semantic object reasoning and spatial position prediction through a visual language model, and determining the position of a target object; based on the panoramic view of the target waypoint, through the intelligent model with the body, detection and positioning are carried out, and a target object is determined; understanding of a high-level abstract instruction is realized, and a target object and a possible position are automatically reasoned; and the navigation capability of the robot in a complex real environment is obviously improved.

The application discloses a low-code platform variable type automatic inference method and device. The method comprises the following steps: screening a variable with an automatic type identifier in a low-code platform to obtain a target variable; obtaining a default value inference result based on a type inference rule; traversing a structured model corresponding to a business logic and declaration information to obtain a node assignment inference result; obtaining a dependency relationship inference result through a multi-round iteration strategy recursive derivation; integrating the default value inference result, the node assignment inference result and the dependency relationship inference result to determine an initial type of the target variable; storing the initial type based on inference trace information and realizing change notification to finally obtain a target type. Through multi-stage inference, expression type calculation, chain dependency processing and other designs, the application does not need to manually specify the type, supports complex scene inference, maintains type safety, reduces use and maintenance costs, and improves the development efficiency and ease of use of the low-code platform.

The invention discloses an automatic inference method and device for low-code platform variable types. The method comprises the following steps: screening and matching variables with automatic type identifiers in a low-code platform to obtain target variables; obtaining a default value inference result based on the type inference rule; traversing the structured model and the declaration information corresponding to the service logic to obtain a node assignment inference result; recursive derivation is carried out through a multi-round iteration strategy to obtain a dependency relationship inference result; integrating the default inference result, the node assignment inference result and the dependency inference result, and determining the initial type of the target variable; and storing the initial type based on the deduced traceability information and realizing change notification, and finally obtaining a target type. Through multi-stage inference, expression type calculation, chain dependency processing and other designs, a user does not need to manually specify a type, complex scene inference is supported, type safety is kept, the use and maintenance cost is reduced, and the development efficiency and usability of a low-code platform are improved.

The invention discloses a workflow automatic production system and method based on natural language analysis, and the system and method enable a workflow building process not to depend on manual node dragging and tedious configuration through the automatic analysis of the natural language input of a user and the intelligent generation of a flow structure. The capability of constructing a complete process without a professional technical background is realized, so that the learning cost is effectively reduced. By means of automatic inference of a deep learning model on a process structure, a process generation process is completed by an automatic model. The system combines preset process knowledge and rules when generating the process, so that standardization of node selection and a connection relationship is realized, structural errors which are easy to occur in manual construction are reduced, and consistency and integrity of process construction are ensured. Besides, through an automatic process conversion and execution monitoring mechanism, the generated process can be rapidly deployed and error analysis and correction can be carried out in the operation process, so that the response capability of the system to the business demand change is improved and the execution maintenance cost is reduced.

This invention belongs to the field of indoor positioning technology and discloses a method and system for constructing and maintaining a long-term WiFi fingerprint map. The method includes: deploying reference points at intervals indoors and collecting WiFi signal strength (RSS) data, coordinates, and timestamps; automatically inferring access point (AP) coordinates based on signal backpropagation and maximum a posteriori probability (MAP) estimation, combined with a path loss model; decomposing the RSS into the mean path loss and residual terms, and modeling the residuals using spatiotemporal Gaussian process regression (GPR); incrementally updating the fingerprint map using an induced point set based on a sparse variational Gaussian process (SVGP); and performing terminal positioning based on the updated map. The system includes five functional units. This invention achieves automatic AP location inference, resistance to dynamic environmental interference, and low-cost long-term map maintenance, improving positioning accuracy and stability.

An operating system state root cause inference method, device, equipment, medium and product are provided. The operating system state root cause inference method comprises: obtaining a structured event with a predefined format in an operating system kernel; determining whether the operating system is abnormally stopped; in response to determining that the operating system is abnormally stopped, obtaining, by a user state service, state information of the operating system based on the structured event obtained in an operating system running cycle before abnormal stop through a predetermined state machine; and obtaining, by the user state service, an abnormal stop root cause based on the state information of the operating system through the predetermined state machine. Through the present disclosure, the structured degree and machine analyzability of operating system abnormal stop analysis can be significantly improved, different fault types can be more accurately distinguished, kernel invasiveness and maintenance costs can be reduced, the implementability of rule iteration and product deployment can be improved, and automatic inference and large-scale review can be realized.

Embodiments of the invention disclose a data processing method and apparatus, a device, a storage medium and a program product. The method comprises the steps of obtaining first behavior sequence data of a target object in a first time period; performing causal relationship learning on the first behavior sequence data through a causal attention mechanism to obtain a behavior causal relationship and a behavior demand relationship of the target object; acquiring second behavior sequence data of the target object in a second time period, and performing reverse causal reasoning on the target object based on the second behavior sequence data and the behavior causal relationship to obtain predicted demand probability distribution of the target object; the first time period is earlier than the second time period; and performing reverse demand reasoning on the target object based on the prediction demand probability distribution and the behavior demand relationship to obtain a prediction behavior of the target object. By adopting the embodiment of the invention, the causal relationship and the deep demand can be automatically deduced from the user behavior, and more accurate user understanding and prediction are realized, so that the user guiding capability is effectively improved.

The invention discloses a code completion model member inference method based on antagonism prompts, which comprises the following steps of: firstly, designing and applying a series of antagonism prompts which keep program functions and are specific to codes, and converting an original input code prefix into a group of disturbed input codes; then querying the target model to respectively obtain prediction outputs of the target model for the original input and the group of input after disturbance, and constructing a feature vector by quantifying the similarity and confusion difference between the two prediction outputs and the real completion code; and finally, training a binary classifier by using the feature vector to automatically deduce the member identity of the sample. According to the method, an inference framework for exposing model memory by utilizing'behavior stability difference 'induced by adversarial prompts is proposed for the first time, an expensive shadow model does not need to be trained, a fine memory mode of a code large language model can be captured, and the accuracy, universality and application feasibility of member inference are remarkably improved.

The invention discloses a power distribution network load margin evaluation method, which comprises the following steps of: automatically deducing a soft adjacency matrix according to node characteristics implied in data through topology deduction analysis without depending on a preset network topology structure by taking voltage amplitude, a voltage phase angle, an active load and a reactive load four-dimensional electrical characteristic of a node as input; modeling a dynamic topology connection relationship of the power distribution network; on this basis, a soft attention GAT mechanism is adopted to extract graph structure features (graph-level features), and node-level representations are aggregated into system-level graph representations through global average pooling; and finally, multi-task collaborative learning is realized through a double-branch output architecture: the LM prediction head carries out supervised learning on labeled data, the feature recovery head carries out self-supervised learning on all data through mask reconstruction, and the LM prediction head and the feature recovery head share bottom layer representation and are jointly optimized. Therefore, accurate and real-time evaluation of the power distribution network LM is realized under the condition that only a small number of labeled data samples are needed.

This invention discloses an automatic inference code generation method and system for model deployment, belonging to the field of model deployment technology. The method includes: extracting model information from intermediate files of the model to be deployed; defining variables for each operator; generating an output variable name index table and a weight variable name index table; determining the input variable names for each operator and generating an input variable name index table according to the order of the operators; generating a pad value index table, an input size index table, and an output size index table for actual inference; for each operator, extracting the corresponding API from an API library; filling the API parameters according to the output variable name index table, weight variable name index table, input size index table, output size index table, input variable name index table, and pad value index table to generate the corresponding operator's API call; and combining the API calls of each operator to obtain the required inference code. This solution provides an automatic inference code generation method for AI chips, simplifying the model deployment process.

The invention relates to the technical field of computer network security, in particular to a network attack traceability evidence obtaining method. Comprising the following steps: firstly, fusing data from various heterogeneous data sources such as malicious software, network traffic, host logs and threat intelligence; then, deep feature extraction is carried out on the multi-modal data through parallel special artificial intelligence models such as a convolutional neural network CNN and a recurrent neural network RNN; then, constructing a heterogeneous attribution knowledge graph by the extracted entities and relationships, and weighting edges in the graph according to credibility and uniqueness of evidences so as to resist deceptive tactics; and finally, analyzing the knowledge graph by using a graph neural network, and automatically deducing the most possible source of the attack by learning high-order complex association among entities in the graph. According to the method, the automation degree, the accuracy and the anti-cheating capability of network attack traceability are remarkably improved.