19 results about "Cognitive engine" patented technology

The invention discloses an artificial intelligence service consultation method based on big data analysis, and the method comprises the following steps: S1, accessing text, voice, image and time sequence data through a multi-source heterogeneous data pipeline, and associating cross-modal data of the same event through a dynamic alignment module; s2, performing multi-modal feature fusion on the data processed in the step S1 by adopting a cross-modal fusion cognitive engine to generate a unified semantic representation vector; and S3, constructing a space-time dynamic knowledge graph based on the semantic representation vector, and detecting an association path and an abnormal mode between the service entities through a graph neural network. According to the method, dynamic alignment and fusion of cross-modal data can be realized through a multi-source heterogeneous data pipeline, business deep association is mined by using a knowledge graph and a causal inference model, and a closed-loop iteration system from quality evaluation to strategy optimization is constructed in combination with reinforcement learning; the core problems of inaccurate demand understanding, insufficient suggestion reliability, low optimization efficiency and the like in traditional consultation are 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.

This invention relates to an intelligent cognitive network control method and system based on business, security, and auditing, belonging to the field of network information technology. The system deploys three cognitive engines—business, security, and auditing—in parallel on the cognitive plane. The method includes: the auditing cognitive engine collecting and analyzing the entire operation and maintenance log in real time to construct an operation chain relationship graph; performing anomaly detection and compliance verification based on an auditing knowledge graph; locating the source of the operation within 10 seconds using indexing and graph technology when an anomaly is detected; and finally driving the self-evolution of auditing knowledge through a feedback loop. This invention achieves deep synergy between performance optimization, real-time defense, and intelligent auditing in network operation and maintenance, solving the technical problems of slow traditional auditing traceability, high compliance costs, and system silos, significantly improving the transparency, security, and autonomy of network operation and maintenance.

The invention relates to the technical field of national defense, in particular to a multi-sensor collaborative national defense electronic security monitoring system which comprises a sensing network layer, a collaborative perception layer, a control decision layer, a communication facility layer and a countermeasure effect layer. According to the multi-sensor collaborative national defense electronic safety monitoring system, multi-source data of radar, photoelectric and infrared thermal imaging, acoustics and vibration sensing and the like are integrated, multi-dimensional features of electromagnetism, vision, acoustics and the like are covered, complex targets are recognized, false alarms are filtered, and the multi-sensor collaborative national defense electronic safety monitoring system is obtained through the regional fusion module and the central situation fusion module. The method comprises the following steps: constructing a global battlefield situation map from sensor data, reasoning the intention of an enemy in combination with a cognitive engine, automatically adjusting sensor parameters, integrating military optical fibers, satellite communication, ad hoc networks and the like according to environment and task requirements, ensuring data transmission under extreme conditions, and integrating physical interception, electronic warfare and network confrontation through a confrontation effect layer. And a closed-loop logic scheme of detection-decision-strike-evaluation is realized.

The invention discloses a power distribution network black-box adversarial sample generation method based on large language model topology reasoning guidance, and the method comprises the following steps: determining an attack target of a scheduling task, quantifying the destruction intention of an attacker, and converting the fuzzy attack target into a multi-target attack loss function; screening out a key node with the highest attack cost performance by utilizing general knowledge and inference capability of a large language model, and obtaining a mask vector; generating a specific disturbance vector for a continuous domain or a discrete domain under the constraint of a mask vector; and performing sample verification. According to the method, a large language model is introduced as a topology cognitive engine, key fragile nodes in the power distribution network are screened out by utilizing the semantic reasoning capability of the large language model, an attack mask is generated, a black box optimizer is guided to lock a key subspace through the mask, and the key subspace is locked under the condition that the internal structure of the model is not accessed. High-quality confrontation samples covering multiple types of attack targets and action domain types are efficiently generated, and rich robustness training data are provided for intelligent agents.

The application discloses a power distribution network black box confrontation sample generation method based on a large language model topology reasoning guide, which comprises the following steps: determining an attack target of a scheduling task, quantifying a destructive intention of an attacker, and converting a fuzzy attack target into a multi-target attack loss function; using general knowledge and reasoning ability of the large language model, screening out key nodes with the highest attack cost performance, and obtaining a mask vector; under the constraint of the mask vector, generating a specific disturbance vector for a continuous domain or a discrete domain; and performing sample verification. The application introduces a large language model as a topology cognition engine, uses semantic reasoning ability of the large language model to screen out key fragile nodes in a power distribution network, generates an attack mask, and guides a black box optimizer to lock a key subspace through the mask, so that high-quality confrontation samples covering multiple attack targets and action domain types are efficiently generated without accessing an internal structure of the model, and rich robustness training data is provided for intelligent agents.

This invention discloses a memory and self-evolution method and system for cognitive engines, belonging to the field of artificial intelligence and machine learning technology. The method includes: storing experience records after each cognitive decision; retrieving relevant historical experiences and modulating candidate situation evaluation values; fine-tuning the polarity of unlocked events online based on cognitive certainty; coordinating polarity based on event co-occurrence statistics; and initializing polarity for new events. This invention endows cognitive engines with experience memory and reuse capabilities, enabling secure online learning driven by cognitive certainty, automatically discovering structured knowledge from event co-occurrence, and protecting core knowledge anchors by locking events. This invention can be added as an independent component to various cognitive engines, giving them life-like self-growth, self-adaptation, and self-repair capabilities.

The invention discloses an embedded path evolution tracking system and an implementation method thereof, and aims to solve the technical problems of incomplete path record, difficulty in problem positioning and insufficient experience inheritance in innovation process management. The system is embedded in innovation platforms such as DIKWP-TRIZ and comprises a path data acquisition module, a log storage module, an organization visualization module, a path analysis module, an improvement suggestion module and a learning optimization module. Key events such as cognitive engine reasoning jump, TRIZ principle application and user decision are intercepted in real time, standardized path units are generated and stored according to a time sequence, and a multi-level evolution path chain and a visual network diagram are constructed. The path analysis module can diagnose and infer breaking points and unresolved contradictions, the improvement suggestion module prompts introduction of new knowledge or adjustment strategies according to the contradictions, and the learning optimization module dynamically optimizes a scheduling algorithm based on a historical data training model. According to the method, full-path transparent recording and intelligent analysis of the innovation process are achieved, efficient redisk, accurate problem positioning and strategy optimization are supported, the innovation success rate and knowledge management efficiency are remarkably improved, and the method is suitable for the fields of research and development management, patent mining and the like.

The application discloses a dynamic interaction cognition and motion control method and system based on a topological field and a medium, and belongs to the technical field of autonomous motion control. The application constructs an interaction cognition engine, and unifies environment perception, interaction cognition and motion control into an endogenous closed loop: endogenous construction of a dynamic interaction topological graph with objects as nodes and normalized interaction risks as edge weights by real-time collection of perception data; generation of a global topological potential field by superimposing three types of potential fields of tasks, safety and interaction, wherein the interaction potential field is directly generated by superimposing the topological graph edge weight and the risk point; generation of a motion trajectory along the negative gradient of the potential field, projection combined with dynamic constraints; and realization of the collaborative evolution of the three by reverse correction of the topological graph and the potential field parameters through motion deviation. The decision-making process is endogenous and interpretable, the algorithmic power requirement is low, the obstacle avoidance success rate reaches 99.9%, and the system can safely operate in a complex dynamic environment without external rules, and is suitable for autonomous navigation and obstacle avoidance scenes of various mobile platforms and dynamic interaction systems.

The invention relates to the technical field of data security, and discloses a financial big data security management platform and a working method thereof, and the platform comprises a collaborative strategy center module which generates a unified collaborative control strategy and issues the collaborative control strategy to other modules; the privacy calculation and analysis module is used for generating cross-domain security insight data based on the strategy; the hidden verification service module is used for generating a ciphertext verification assertion based on the strategy; the intelligent cognitive engine module is used for fusing the cross-domain security insight data, the ciphertext verification assertion and the real-time security data flow and generating comprehensive security situation information; and the full-link evidence storage module captures key operation and output of each module and forms a tamper-proof full-life-cycle auditing evidence chain. The method corresponds to the platform. Through cooperation of strategy driving and multiple modules, cross-domain security management and credible auditing of financial data under privacy protection are realized.

The invention belongs to the technical field of power system informatization and intelligentization, and particularly relates to a power transmission professional autonomous intelligent office system and method based on multi-modal perception, digital twinning and artificial intelligence technologies, and the system comprises a cross-modal business model which serves as a core cognitive engine of the system; the multi-mode fusion sensing module is used for comprehensively judging the intention of a user and the state of an office scene through voiceprint, voice, gestures and environment sensor data; the power transmission line digital twinborn body is synchronized with the power transmission line of the physical world and the panoramic platform data in real time, and is used for state prediction, fault simulation and operation deduction; the workflow engine with intelligent driving autonomously plans, schedules and executes a complex office task sequence, and the system can realize the crossing from passive response to active prediction and from process solidification to intelligent adaptation.

The invention provides a system for optimizing communication between AI agents, which can respond to requests related to tasks and process and explain real-time scene data. The system comprises one or more sensing devices used for collecting scene data in real time according to a request of a second AI agent; the large language model module is used for generating a task-specific event dictionary when the request is related to the task; the sensing engine is used for processing the collected data; the cognitive engine is used for extracting semantic information; and the decision engine is used for identifying events related to the task according to the event dictionary. An operation platform of the system comprises a communication module, a processor and a memory, and when stored instructions are executed, the system can run among a perception layer, a cognitive layer and a decision layer. According to the invention, context-based hierarchical information exchange is ensured, so that the task requirements of the AI intelligent agent are met.

The invention provides an Internet of Things Agent adaptive interaction system driven by a multi-modal large model, and relates to the technical field of data processing, and the system comprises a primitive generation module which is used for inputting a joint semantic embedding tensor into a pre-trained Internet of Things Agent cognitive engine driven by the large model to obtain a hidden state evolution sequence, and performing causal reasoning and compliance constraint decoding on the hidden state evolution sequence to generate a structured interaction strategy primitive. Through cooperation of the modules, adaptive generation, dynamic optimization and execution compliance verification of the Internet of Things Agent interaction strategy are realized, the precision of cooperative control of the Internet of Things equipment is improved, and multi-dimensional quantitative evaluation can be performed on the execution effect at the same time.

This application discloses a reasoning method, apparatus, electronic device, and storage medium based on a dual-engine separation architecture. The method and apparatus are applied to an electronic device. Specifically, in response to a user's input request, an original high-resolution image is input into a perception engine. The perception engine independently performs key region localization operations on the original high-resolution image, outputting at least one key region, its location information, and a corresponding confidence score. It then determines whether the confidence score meets a preset trigger condition. If so, based on the location information of the key region, a corresponding high-resolution sub-image is cropped from the original high-resolution image. The high-resolution sub-image and a natural language query command are input into a cognitive engine. The cognitive engine performs a self-refining reasoning process based on the high-resolution sub-image, including at least an initial judgment stage and an evidence verification stage, and outputs the final semantic judgment result. This application, through a perception-cognition separation architecture design, combined with conditional triggering scheduling and multi-round self-refining reasoning, fundamentally solves the problem of not being able to perceive the microscopic visual evidence required to discern key behaviors in high-resolution image understanding, thus improving the accuracy of recognizing concealed and fine-grained behaviors.

The application provides a multimodal large model driven Internet of Things Agent adaptive interaction system, relates to the technical field of data processing, and comprises a base element generation module configured to input a joint semantic embedding tensor into a pre-trained large model driven Internet of Things Agent cognitive engine, obtain a hidden state evolution sequence, and perform causal reasoning and compliance constraint decoding on the hidden state evolution sequence to generate a structured interaction strategy base element. Through the cooperation of various modules, the application realizes the adaptive generation, dynamic optimization and execution compliance verification of the Internet of Things Agent interaction strategy, improves the accuracy of Internet of Things device collaborative control, and can perform multi-dimensional quantitative evaluation on the execution effectiveness.

The invention relates to a security endogenous and cognitive self-evolution intelligent network control method and system, and belongs to the technical field of network information. According to the system, in a traditional four-plane unit architecture, a security cognition engine parallel to a service cognition engine is innovatively introduced into a cognition plane unit. The method comprises the following steps: constructing a user behavior baseline portrait through unsupervised learning; comparing the user operation behavior with the baseline portrait in real time, and calculating a behavior deviation degree; an access control strategy is generated in a self-adaptive mode based on the deviation degree, and real-time non-inductive blocking of abnormal access is achieved; and finally, driving self-evolution of the security knowledge by executing feedback. According to the method, the continuous verification technology based on the user behavior portrait is deeply integrated into the network self-evolution closed loop, internal biochemistry, self-adaption and self-evolution of the security capability are achieved, the technical problems that traditional network security protection lags behind, experience is poor, and operation and maintenance are complex are effectively solved, and high-order unification of security and efficiency is achieved.

The invention provides a bank virtual digital human service system and method fused with multi-modal interaction. The invention discloses a bank virtual digital human service system fused with multi-modal interaction. The system comprises a multi-modal data acquisition end; edge computing nodes; the cloud cognitive engine is provided with a financial intention-emotion coupling perception network and is configured to extract semantic features, visual features and audio emotion features through a text encoder, a visual encoder and an audio encoder respectively; retrieving related financial entity concepts according to the semantic features, and generating knowledge embedding vectors; performing multi-head attention calculation on the knowledge embedding vector with the visual features and the audio emotion features by using a knowledge graph enhanced cross attention module, and outputting a user intention vector and an emotion load index; and the man-machine cooperative control center is provided with a dynamic risk gating module. According to the method, the service capability and the security of the bank virtual digital human in a complex business scene are remarkably improved.

The application discloses an education cognition engine adaptive optimization method based on a large model, which comprises the following steps: S1, collecting teaching behavior data from an education database according to collection specifications, forming a unified format through an ETL tool, and mapping the behavior data into a cognitive state vector by combining a preset cognitive dimension through a cognitive signal encoder; S2, extracting cognitive correlations from the cognitive state vector through BERT, and obtaining an evolutionary knowledge graph with evolutionary markers through an evolutionary marker mechanism optimized by cognitive correlation weights; S3, based on the evolutionary knowledge graph, introducing an improved graph neural network to model cognitive evolution links and obtain a cognitive link evolution model, and generating an optimal candidate strategy list through the cognitive link evolution model; and S4, based on the optimal candidate strategy list, accessing a basic education cognition engine to automatically match teaching resources, effectively meeting the learning needs of different students, and improving the pertinence and effectiveness of teaching.

The invention provides an enterprise cloud data autonomous governance system supported by a polymorphic cognitive engine, which relates to the technical field of enterprise data systems, is deployed in an enterprise-level hybrid cloud environment and is composed of a data acquisition module, a governance engine module, a storage module, a monitoring module, a privacy adaptive module, a sustainability evaluation module and a distributed consensus unit. According to the system, structured, unstructured and real-time streaming data are acquired in a multi-modal acquisition mode, and heterogeneous data are converted into a pulse sequence which can be identified by a neuromorphic processing unit through a pulse precoding unit. A multi-state cognitive engine is arranged in the governance engine, the governance engine is composed of a pulse neural network structure and a normal form switching mechanism, and self-iteration optimization can be achieved through generative sample construction and evolution parameter updating according to the statistical characteristics of a pulse mode. And the storage module expresses a complex business relationship by adopting an associated storage mode based on a hypergraph structure. According to the invention, autonomy, structuralization and credibility of data governance in the enterprise cloud environment are realized.