303 results about "Inference system" patented technology

The invention belongs to the field of data processing, and particularly relates to a dynamic cooperative control system and method for a gas turbine and a micro-grid, and the method comprises the steps: constructing a micro-grid real-time monitoring module, continuously collecting distributed energy real-time output, controllable load demands, bus voltage frequency and equipment state parameters, and carrying out the filtering and noise reduction through a preprocessing unit, thereby guaranteeing the data precision; calculating a real-time power difference value based on the preprocessed data, calling an adaptive neural fuzzy inference system, taking the power difference value, the bus voltage deviation and the frequency deviation as input, and judging whether the power difference value, the bus voltage deviation and the frequency deviation exceed a preset threshold value by means of a fuzzy rule base and a neural network model; if the threshold values are not exceeded, the current states of the gas turbine and the energy storage system are maintained; if any one exceeds the threshold value, a dynamic cooperative control instruction is triggered, precise cooperative control of the gas turbine and the micro-grid is achieved, and the operation stability, the operation efficiency and the reliability of the micro-grid are improved.

A database structure obtained by means of information retrieval, and a reasoning system, which structure and system specifically relate to a gut microbe knowledge graph system, comprising: a gut microbe knowledge graph consisting of a gut microbe knowledge base, a gut microbe and small-molecule drug therapy association knowledge base, and a clinical medicine database; and a multimodal uncertainty reasoning system, using the gut microbe knowledge graph. The gut microbe knowledge graph system predicts potential diseases, drugs, genes, etc., which are associated with gut microbes.

The invention belongs to the technical field of artificial intelligence, and relates to an artificial intelligence-based electric energy quality disturbance identification and positioning method, which comprises the steps of constructing an electric energy quality disturbance signal data set, performing segmented preprocessing on electric energy quality disturbance voltage data, enhancing time-frequency joint features and encoding disturbance sensitive areas. And constructing a deep learning model for power quality disturbance identification and positioning, and identifying and positioning the power quality disturbance. According to the invention, through adaptive denoising processing, boundary detection and multi-resolution time-frequency feature extraction, the identification precision and positioning precision of power quality disturbance are significantly improved; self-adaptive wavelet denoising and dynamic segmentation are combined, noise interference is effectively suppressed, and the edge characteristics of voltage sudden change points are kept; according to the dual-task sharing network, disturbance identification and positioning tasks are cooperatively optimized, so that the network can consider disturbance classification and time positioning at the same time; and through Bayesian reasoning, the system can output confidence estimation, provides credibility quantification of identification and positioning results, and effectively improves the reliability of the system.

The invention discloses a large language model end-cloud collaborative inference system based on low-rank fine tuning, and belongs to the technical field of inference optimization of end-side cloud computing. Establishing an end-cloud collaborative reasoning architecture, and in an offline stage, performing parameter fine tuning on a large language model by a cloud side based on training data of different downstream tasks; in the online stage, user requests are classified through'variational auto-encoder-Gaussian mixture model 'clustering, whether a low-rank adapter matched with a current task exists in an end side cache is judged, and if yes, reasoning is executed on the end side; and otherwise, forwarding the task to the cloud side. After a plurality of user requests are processed by the architecture, historical user requests and cache states are analyzed based on a Mama model, and an end-side low-rank adapter library is dynamically updated. And monitoring end cloud load and reasoning delay in real time, and issuing the new adapter to the end side according to the task repetition rate increment. According to the method, dynamic balance of the system is realized, and high efficiency and adaptability of the system are ensured while calculation and storage overhead are reduced.

The invention discloses a multi-agent large-model disease diagnosis knowledge reasoning system based on data dual drive, and relates to the technical field of artificial intelligence assisted medical diagnosis. The system collects patient symptom follow-up records, laboratory test results, observation diagnosis probabilities and expert diagnosis recommendation results in a multi-source manner; time sequence evolution characteristics are extracted, a time sequence diagnosis sensitivity coefficient is calculated, and early recognition of disease risks is achieved; in combination with anti-fact simulation and statistical reasoning, a causal consistency coefficient is obtained and is used for verifying causal reasonability of observation diagnosis and contrast results; based on agent group consensus analysis, calculating a game consistency coefficient for judging the credibility of a diagnosis conclusion; positioning and multi-level verification are carried out on abnormal reasoning steps and knowledge fragments, so that the reliability and safety of a result are guaranteed; continuous optimization of the diagnosis model is realized through a log analysis and knowledge backflow mechanism; according to the invention, the accuracy, interpretability and safety of disease diagnosis can be obviously improved.

The invention discloses a memory limited device MoE large model reasoning optimization system and method based on dual prediction, and belongs to the technical field of MoE large model reasoning optimization. In order to solve the problem of optimizing the reasoning performance of the LLM based on the MoE on the memory limited equipment, the system comprises a layer-by-layer predictor, a token-by-token predictor and an expert management module, and the expert management module comprises an expert cache, a temporary expert buffer area and an I / O scheduler. The layer-by-layer predictor is connected with the token-by-token predictor, the layer-by-layer predictor and the token-by-token predictor are respectively connected with the I / O scheduler, the token-by-token predictor is connected with the expert cache, and the I / O scheduler is respectively connected with the temporary expert buffer area and the expert cache. According to the method, a predictive expert Cache, a temporary expert buffer area and a token-by-token prefetching technology are innovatively provided; according to the invention, the reasoning system can use fewer memory resources, and the LLM reasoning speed is obviously improved.

The invention relates to the technical field of network security, and provides an APT attack traceability and path restoration method and system, attack path validity is verified through a dynamic causal element path generator in combination with an anti-factual adversarial network, and fine-grained entity modeling is realized through a hierarchical multi-modal entity inference system. A causal attention mechanism is optimized to improve association analysis precision, a federal incremental learning framework is constructed to realize dynamic updating and privacy protection, and the problems of poor rule adaptability, coarse granularity of entity modeling, insufficient causal association distinguishing and low calculation efficiency in the traditional technology are effectively solved. The method has the technical effects of dynamically generating an effective attack path, improving attack traceability accuracy, reducing calculation overhead and enhancing privacy protection.

Various examples, systems, and methods are disclosed relating to an agentic curation pipeline. One system can process questions and other inquiries about video content by using a combination of models and stored information. The system can receive a query related to an event in a video, selects relevant portions of the video using embeddings, and apply the selected video data and a related sub-query to a video model. The output from the video model can be used by a language model, along with stored context, to generate an answer to the original query. The system can returns the answer to the requester.

The invention belongs to the technical field of intelligent question-answering systems, and particularly relates to an intelligent question-answering inference system based on a knowledge graph, which is characterized in that firstly, a knowledge graph construction module fuses multi-source data to generate a structured graph, and after a user inputs a natural language question, a question-answering analysis module completes intention classification and entity disambiguation and converts the question into structured query; an inference engine module fuses symbol rules and graph neural network inference through a hybrid inference sub-module, and a dynamic weight adjustment sub-module optimizes weights according to errors and attenuation factors to generate an inference result; the knowledge updating module incrementally updates the atlas in real time and detects conflicts, the interactive interface module visually presents a result, and the evaluation optimization module iteratively optimizes parameters in combination with offline evaluation and online feedback. The whole process is from user question asking to result output, accurate reasoning and continuous performance improvement are achieved, and multi-field question and answer requirements are met.

The invention relates to the technical field of network security and data processing, and discloses a violation short message recognition method and system based on deep semantic understanding, and the method comprises the steps: firstly cleaning an original short message, generating a mixed embedding vector through characters, sub-words and pinyin, and carrying out the recognition of the violation short message; then processing through a double-layer detection engine, wherein the first layer utilizes rules and a lightweight model for rapid preliminary screening; in the second layer, for suspected samples, a double-tower fusion neural network architecture is adopted, local and global features are combined, fusion is carried out through a gating unit, and a large language model is input to carry out deep semantic reasoning. The system executes strategies such as interception or flow limiting according to the risk score, and realizes model iteration through a dynamic knowledge base and incremental learning. According to the method, the resource consumption and the detection precision are balanced through the layered architecture, the antagonistic variants are effectively identified by utilizing multi-dimensional feature fusion, and the method has the adaptive evolution capability for a novel violation mode.

The invention discloses an industrial equipment fault inference system based on a knowledge graph, and the system comprises a knowledge graph construction module, a fault data collection module, an inference analysis module and a response processing module. The entity extraction unit extracts equipment components, fault types and maintenance record entities from an industrial equipment operation document, equipment manual unstructured data supplementation attributes are integrated, the attributes and association weights are marked, and the relationship construction unit establishes a fault causal relationship between the entities and a component association relationship to form a multi-level knowledge network; a knowledge verification unit verifies entity attribute consistency and relation rationality, a dynamic updating unit receives data updating nodes and relation strength of each unit and receives feedback data optimization weights, and in a fault data acquisition module, a real-time monitoring unit acquires operation parameters and state signals and associates equipment identifiers.

The invention relates to the technical field of financial knowledge management, and discloses a multi-modal Agent coordinated reasoning system, and the key points of the technical scheme are that the multi-modal Agent coordinated reasoning system comprises a data interface module which is used for connecting a data source to obtain target data, and is also used for receiving a task instruction; the function management module is used for constructing functional Agents with various data processing functions and executing corresponding data processing on target data according to a scheduling strategy through the functional Agents; the business management module is used for constructing a business Agent of each business field and carrying out business processing according to a scheduling strategy through the business Agent; and the scheduling decision module is used for generating a scheduling strategy according to the task instruction and the target data, and performing scheduling and collaboration on the function Agent and the service Agent according to the scheduling strategy.

The invention belongs to the technical field of thinking chain reasoning, and particularly relates to a multi-modal reasoning method and system based on error attribution. The reasoning method comprises the following steps: on the basis of a current modal fusion weight, performing modal fusion on each piece of initial information in an initial information set, and then generating a thinking chain; after a reasoning dependency graph is constructed based on the thinking chain, check points are selected in the reasoning dependency graph; based on consistency, factuality and logicality, performing error possibility scoring on each check point, and if the error possibility scores of all check points in the current thinking chain are below a set threshold, outputting the current thinking chain; otherwise, marking the check points of which the error possibility scores exceed a set threshold value as error nodes; calculating relative contribution strength of different modes to error nodes; and on the basis of the relative contribution strength, updating the modal fusion weight, and regenerating the thinking chain. According to the invention, the accuracy of the reasoning result and the stability of the accuracy can be improved.

The invention discloses a large-model adaptive batch reasoning system based on time delay measurement, and the system comprises a request receiving and queue management module which is used for monitoring and receiving an LLM reasoning request sent by an external user or an application; the real-time delay sensing and batch monitoring module is used for measuring and analyzing performance indexes directly related to the current batch processing strategy in real time in the LLM reasoning process; the Token-by-Token batch processing and reasoning execution module is used for selecting a corresponding number of requests from the request queue according to the size of the current batch determined by the scheduler, organizing the requests into an effective calculation batch and submitting the effective calculation batch to the LLM reasoning core at the bottom layer to execute one or more decoding steps; and the double-stage batch adjustment decision module is used for obtaining the normalized time delay ratio. The invention also discloses a large-model adaptive batch reasoning method based on time delay measurement. According to the invention, substantive improvement and intelligent management of the overall performance of the LLM reasoning service are realized.

The invention provides a knowledge enhancement type sports video understanding method based on large-model dual-mode reasoning, and belongs to the field of video understanding. Firstly, a sports video needing questioning and a question text are obtained, and the sports video, the question text and cue words are input into a reactive reasoning agent; the reactive reasoning agent classifies the questions according to the question texts and the cue words, and if the questions belong to simple questions, the reactive reasoning agent answers the questions according to the input sports videos; if the questions belong to complex questions, the questions are answered through a deep reasoning agent, and the deep reasoning agent is composed of a dynamic motion divider, a key segment selector and a fine-grained matcher based on the sports knowledge graph. According to the method, a dual-mode reasoning system is innovatively introduced, the dynamics and domain specificity of the sports video and the diversity and complexity of questions asked by the user are fully considered, and the performance of a (multi-mode) large language model in a sports video understanding task is remarkably improved.

A variety of smart inference systems and robotic control devices are disclosed. A device networking system interconnects mobile devices, robotic devices, and others. A robotic controller secures and manipulates a controlled device user interface. A smart device system interprets data based on a plurality of models operating on a plurality of runtimes based on a plurality7 of capabilities. A smart payment processor system assigns payments accounts automatically as a function of a semantic matching between one or more inferred semantic identities. A robotic emulation device captures and analyzes a video signal from a target device and transmits manipulation signals emulating a peripheral input device. A system of carts are physically coupled in a charging configuration. A conveyor is configured to couple with another conveyor and having sensors to detect weight, size, location or other aspects of conveyed items, and implementing semantic analysis for management and augmentation.

The invention provides an inference system and method of a large language model based on multi-chip parallel computing, and relates to the technical field of artificial intelligence. The system comprises a pre-calculation module used for processing input instruction information to generate to-be-reasoned data, and the to-be-reasoned data is in a matrix form; the expert parallel module is used for sending the to-be-reasoned data to accelerator chips in the expert parallel module and determining sub-reasoning data processed by the activation expert units corresponding to the accelerator chips respectively, so that the activation expert units carry out calculation based on the corresponding sub-reasoning data and complete parallel calculation result data is determined. The input data is broadcasted to all the accelerator chips, each accelerator chip selects the corresponding input data for calculation according to the set activation expert unit, the same complete calculation result is obtained through global protocol operation among all the accelerator chips, and the overall operation performance and efficiency are improved.

The invention discloses a multi-agent collaborative knowledge reasoning system based on a large language model, and relates to the technical field of intelligent manufacturing and artificial intelligence. Natural language output of a large language model is converted into rules, facts and ontology fragments which can be directly consumed by an inference engine through knowledge obtaining and compiling, continuous increment updating of cross-domain knowledge is achieved in cooperation with metadata with sources and timestamps, and the limitation that a traditional static knowledge base is difficult to cover dynamic faults is overcome; secondly, a blackboard and agenda mechanism is used as a cooperative carrier, intermediate assertions of intelligent agents such as vibration, circuits and logs are published and subscribed in a structured mode, and a conflict resolution and consistency verification module carries out unified judgment according to specificity, time freshness and source credibility; therefore, delay and uncertain accumulation caused by long-chain natural language dialogues are avoided in a strong real-time scene.

The invention discloses an inference method and an inference system of a gene regulatory network. The inference method comprises the following steps: acquiring transcriptome data and prior gene network data of a single cell; extracting a gene sub-network related to the transcriptome data from the prior gene network data; obtaining a first view and a second view for the gene sub-network according to a first random deletion strategy and a second random deletion strategy; providing the first view and the second view to a neural network model, and performing comparative learning based on the neural network model to obtain incoming features and outgoing features of each node in the gene sub-network; and constructing a regulation score matrix based on the incoming features and the outgoing features, wherein the regulation score matrix displays the regulation association degree between genes in the transcriptome data of the single cell. According to the technical scheme, the direct causal relationship and the indirect association relationship can be effectively distinguished, so that the gene co-expression network more accurately reflects the real regulation relationship.

The invention relates to the technical field of artificial intelligence, in particular to a browser-side AI reasoning system and method based on TensorFlow.js. The browser-side AI reasoning system and method based on TensorFlow.js comprises the following steps of model loading and initialization, equipment performance detection and model selection, input data collection and preprocessing, reasoning task scheduling and execution, result analysis output and visual export. Model cache updating maintenance and performance monitoring dynamic adjustment and optimization are carried out; the method has the beneficial effects that a set of complete front-end AI reasoning flow control mechanism is provided; realizing model dynamic adaptation based on equipment performance; various input forms are supported; a main thread is prevented from being blocked by utilizing multi-thread scheduling; a GPU acceleration and mixing precision calculation mechanism is introduced to improve the reasoning efficiency; establishing a model caching mechanism to improve the loading speed and the offline availability; performance monitoring and dynamic tuning functions are provided, and reasoning stability is ensured; providing a structured result output and visual display interface; a model version control and background hot update mechanism is realized; and the security, compatibility and expansibility of the system are improved.

The invention discloses a reasoning and interaction system based on multiple dimensions. According to the inference system based on multiple dimensions, the complete process from multi-source data acquisition to inference result output is realized. The acquisition module is responsible for collecting out-hospital examination information and in-hospital examination information of a patient. And the pruning module predicts a targeted set by utilizing the interactively collected patient information and a preset diagnosis knowledge base, dynamically prunes the full-amount question-examination thinking sub-trees according to the targeted set, and optimizes the pruned question-examination thinking sub-trees. And the feature extraction module extracts key features from the interactively collected out-of-hospital and in-hospital question examination information based on the optimized question examination thinking sub-tree. And finally, the inference module inputs the key features into a pre-trained diagnosis model, and the model is combined with a Bayesian algorithm, a deep learning algorithm and a clinical decision consensus to output a final inference result. The problem that dynamic adaptation and efficient and accurate reasoning of multi-dimensional data cannot be achieved on a task set with specific requirements in the prior art is effectively solved.

The invention discloses a big language model credible reasoning system and method based on double-flow cognitive loss and solution synthesis, and aims to solve the problem of'compliance illusion 'caused by the fact that a big language model cannot process internal and external knowledge conflicts. The invention provides a double-flow cognitive engine, which is used for executing a'this-my 'flow of internal knowledge of a simulation model and a'super-my' flow strictly following an external document in parallel so as to generate two opposite cognitive state vectors. The system calculates a semantic conflict score between the two vectors in real time through a cognitive disorder detector. When the conflict is significant, the system activates and demodulates the synthesis module, and actively generates a text for solving the conflict; when there is no conflict, the output of the "super-my" stream is continued to ensure factuality. According to the method, cognitive conflicts are explicitly and actively solved, and a large language model is transformed into an inference engine with inherent critical thinking and high credibility.

The invention discloses a distributed intelligent deduction system and method for multi-source data fusion and dynamic scheduling, and relates to the technical field of deduction simulation. The deduction preprocessing module is used for carrying out task stage division on military scenarios and setting decision points, generating branch tasks according to tactical rules and carrying out model classification; the deduction management module is used for promoting deduction from an initial state based on a simulation engine, generating branch tasks at a decision point according to a real-time state, distributing initial weight coefficients by combining historical efficiency parameters and model types, and realizing parallel deduction and dynamic model switching; the deduction optimization module continuously updates a weight coefficient through a branch efficiency parameter, screens an optimal deduction path and iterates a state snapshot in a closed loop; the resource regulation and control module collects data in real time to support efficiency evaluation, resource distribution is dynamically optimized according to task loads, the deduction efficiency and accuracy are remarkably improved, and intelligent resource scheduling and system self-adaptive optimization are achieved.

A GPU-sharing method and apparatus for serverless inference loads is provided, wherein the method involves intercepting and forwarding GPU API calls made by inference tasks to an API proxy process to manage and allocate GPU resources. With a CPU and multiple GPUs connected through a bus, the GPUs communicate with the CPU only through an API proxy for process management and resource allocation of the GPUs. The CPU intercepts all GPU APIs triggered by any function of a same inference application, forwards the intercepted GPU APIs to a same designated GPU runtime for execution, and directs the GPU APIs triggered by each function to a pre-designated stream pool for the same inference application, so that all the functions of the same inference application share the same GPU runtime. The present disclosure solves the problem related to bulkiness of GPU runtimes in serverless inference systems, thereby facilitating GPU resource usage.

The invention discloses an intelligent feces treatment system and method, and belongs to the field of sewage treatment.The intelligent feces treatment system comprises a data acquisition module used for collecting COD, ammonia nitrogen, DO, pH, water temperature, SS and microbial activity parameters of sewage in real time and conducting normalization processing on the collected water quality parameters to form a unified standard data basis; the first model module is used for optimizing membership function parameters and fuzzy rules of an adaptive neural fuzzy inference system by applying a particle swarm algorithm based on the data processed by the data acquisition module, outputting a predicted value of COD concentration in a biochemical pool in real time by training an ANFIS model, and forming a COD prediction model; and the second model module is used for constructing a Stacking integrated model by taking the COD predicted value in the first model module and the data processed in the data acquisition module as input, the integrated model comprises a base learner and a meta learner, and then a prediction result of the base learner is generated by adopting a five-fold cross validation mode.

The invention provides a multi-model adaptive reasoning system and reasoning method, and is applied to the technical field of artificial intelligence. The system comprises four core modules: a user side is responsible for receiving and forwarding a request; the routing layer module serves as a decision-making center, analyzes request types and intelligently distributes the request types to the reasoning layer; the reasoning layer module executes a specific reasoning task and returns a result; the cache layer module stores a high-frequency request result to improve response efficiency; the implementation process of the method is as follows: submitting a request by a user, analyzing and distributing by a routing layer, processing and caching a result by a reasoning layer, and regularly optimizing a model based on cached data. According to the system, through a dynamic routing and intelligent caching mechanism, the consumption of computing resources is effectively reduced, and the response speed and the processing efficiency of the artificial intelligence service are improved.

An inference system applies a machine-learning transformer model to a batch of requests with variable input length or variable target length or variable internal sate length by selectively batching a subset of operations in the transformer model but processing requests in the batch individually for a subset of operations in the transformer model. In one embodiment, the operation to be processed individually is an attention operation of an encoder or a decoder of the transformer model. By selective batching, the inference system can allow batching operations to be performed for a batch of requests with variable input or target length or internal state length to utilize the parallel computation capabilities of hardware accelerators while preventing unnecessary computations that occur for workarounds that restrain the data of a batch of requests to a same length.

The invention relates to a cloud edge multi-unmanned aerial vehicle cooperative resource optimization method assisted by a large language model, and belongs to the technical field of unmanned aerial vehicle communication, and the method comprises the following steps: S1, constructing an edge-cloud unmanned aerial vehicle cooperative reasoning system; s2, establishing a joint optimization model for discriminating gain maximization; s3, deploying a large language model at a cloud node, and generating a global strategy through a planner, a memory bank and an reflection evaluator; s4, deploying a deep reinforcement learning model at each edge node, and executing real-time optimization according to a global strategy and local observation data; s5, a collaborative feedback mechanism is established, the edge node feeds back an execution result to the cloud node, the cloud node updates a global strategy according to the feedback result, and the edge node adjusts real-time optimization parameters according to the updated global strategy; and S6, adopting an actor-commentator resource allocation algorithm for dynamic knowledge flow collaborative optimization, and realizing collaborative optimization through a distributed sensing and centralized decision framework.

The invention relates to a computer system utilizing a computer model. A method, a computer storage medium, a program product and an apparatus for training an isolated routing model, and a hybrid inference system are disclosed. The method for training the isolation routing model comprises the following steps: acquiring a training task set; obtaining a first model passing rate and a second model passing rate; generating a soft label based on the first and second model passing rates; generating, by the isolated routing model, a prediction result indicating a probability that the training task is to be routed to the first or second major language model; determining a loss function based on the prediction result and the soft label; and adjusting parameters of the isolation routing model based on the loss function. The first and second model passing rates are determined by sampling a plurality of times a result of performing the training task on the first and second large language models, respectively. According to the method, the isolation routing model can be efficiently trained, so that tasks can be accurately routed among a plurality of large language models to obtain balance between performance and cost.

The invention discloses a power taking device dynamic control method based on an intelligent algorithm, and the method comprises the following steps: S1, collecting real-time operation parameters of a power taking device, processing the real-time operation parameters, and generating a time sequence feature data set; s2, inputting the feature data set into a liquid neural network model with an adaptive gating mechanism, and outputting intermediate feature expression; s3, inputting the intermediate feature expression and the symbol rule into a neural symbol inference system, and outputting a control decision instruction of the power taking device; s4, performing comparison processing on the control decision instruction according to a preset equipment safety rule base; and S5, adjusting the power taking device according to the control decision instruction, collecting the adjusted data in real time, and circularly executing the steps S1 to S5 until the working state is terminated. According to the method, the liquid neural network model and the neural symbol inference system are fused, and the adaptive gating mechanism is added in the liquid neural network model as an improvement, so that efficient cooperation and management requirements in a novel intelligent power consumption scene are met.