Brain's autonomous decision-making system
By constructing a brain-based autonomous decision-making system, the problems of the disconnect between algorithm development and training processes and the lack of dynamic monitoring have been solved. This has enabled efficient management of data resources and high precision and adaptability of the decision-making system, thereby improving system synergy and decision reliability.
Patent Information
- Application Number
- CN202611061827.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-16
- Publication Date
- 2026-11-17
AI Technical Summary
Existing decision-making systems struggle to meet the demands for high-precision and highly adaptable decision-making in complex scenarios. The disconnect between algorithm development and training, coupled with a lack of dynamic monitoring and feedback, leads to resource waste and insufficient accuracy in decision-making results.
We construct a brain-based autonomous decision-making system, including a data configuration module, an algorithm factory module, a cognitive inference and computation module, a basic model management module, and a decision output module. This system achieves a tight closed loop of dataset management, algorithm development, and training, and optimizes model deployment and decision output by combining dynamic monitoring and feedback mechanisms.
It enables centralized management and efficient adaptation of various data resources, reduces resource waste, improves system collaboration and decision reliability, and enhances decision accuracy and adaptability through dynamic monitoring and feedback mechanisms.
Smart Images

Figure CN123117316A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of intelligent decision-making and artificial intelligence technology, specifically to a brain-based autonomous decision-making system. Background Technology
[0002] With the rapid development of artificial intelligence technology, intelligent decision-making systems have been widely applied in various fields such as financial risk control, industrial manufacturing, and intelligent transportation. Their core requirement is to achieve autonomous and accurate decision output through the analysis of massive amounts of data, algorithmic computation, and model deduction. Existing decision-making systems mostly adopt a linear architecture of "data input - single algorithm computation - result output." While this can complete basic decision-making tasks, it exposes many shortcomings in complex scenarios, making it difficult to meet the demands for high-precision and highly adaptable decision-making.
[0003] Current decision-making systems generally suffer from the following problems: First, the algorithm development and training processes are disconnected. There is a lack of coordination between algorithm editing, training task creation, and model generation. When algorithm parameters are adjusted, the training environment needs to be reconfigured, and the resources developed in the early stages cannot be directly reused. For example, in a financial risk control system, the optimized risk control algorithm requires rebuilding the training environment, wasting a lot of time and resources. Second, cognitive inference lacks dynamic monitoring and feedback. The inference process of existing systems is mostly a black box operation, which cannot adjust the inference strategy according to real-time training data (such as changes in loss function and resource usage), resulting in insufficient accuracy of decision results. Summary of the Invention
[0004] To address the technical problems of the disconnect between algorithm development and training processes, and the lack of dynamic monitoring and feedback in cognitive inference, this invention provides the following technical solution:
[0005] The brain's autonomous decision-making system includes:
[0006] The data configuration module collects data from multiple sources, classifies, cleans, converts, and labels the data, builds a unified repository and indexes it, and outputs a standardized dataset library with indexes.
[0007] The data security and compliance verification module, based on the standardized dataset library of the data configuration module, configures scenario-based compliance rules, and outputs a security compliance dataset and traceability log after de-identification, security verification and logging.
[0008] The Algorithm Factory module, based on the security and compliance dataset output by the Data Security and Compliance Verification module, completes algorithm development, verification, training configuration and training, and outputs the trained algorithm and real-time training process data.
[0009] The cognitive inference calculation module, based on the trained algorithm and real-time training process data output by the algorithm factory module, configures the scenario, dynamically infers, optimizes and evaluates, and outputs a preliminary inference model that has passed the evaluation.
[0010] The basic model management module, based on the prototype of the inference model output by the cognitive inference calculation module, is encapsulated, optimized, version managed and deployed, and outputs a callable model service interface with version identifier.
[0011] The decision output module, based on the callable model service interface output by the basic model management module and the standardized dataset library of the data configuration module, outputs the final decision conclusion and provides feedback after access, decision invocation, and result verification.
[0012] As a preferred embodiment of the brain-based autonomous decision-making system described in this invention, the data configuration module includes:
[0013] The dataset acquisition unit collects data from multiple sources and outputs the raw dataset.
[0014] The dataset classification unit classifies the raw data set output by the dataset collection unit according to data type and application scenario, and outputs the classified raw data set.
[0015] The dataset processing unit performs data cleaning, format conversion, and label annotation on the raw dataset output by the dataset classification unit, and outputs a standardized dataset.
[0016] The dataset storage and indexing unit builds a unified repository for custom datasets and pre-built datasets based on the standardized datasets output by the dataset processing unit. It also establishes multi-dimensional indexes to output a standardized dataset library with indexes.
[0017] As a preferred embodiment of the brain-based autonomous decision-making system described in this invention, the data security and compliance verification module includes:
[0018] The compliance rule configuration unit, based on the standardized dataset library output by the data configuration module, combines the compliance requirements of different industries to configure data usage permissions, sensitive data identification rules and compliance verification standards, and outputs a scenario-based compliance rule set;
[0019] The data desensitization unit, based on the rule set of the compliance rule configuration unit and the standardized dataset library output by the data configuration module, performs desensitization processing on data containing sensitive information and outputs the desensitized standardized dataset.
[0020] The security verification unit, based on the standardized dataset output by the data anonymization unit, verifies the encryption status of data transmission, access permission adaptability, and anonymization integrity, while also blocking illegal data and outputting a security compliance dataset and verification report.
[0021] The compliance log recording unit records data processing information and outputs log files based on the verification report from the security verification unit.
[0022] As a preferred embodiment of the brain-based autonomous decision-making system described in this invention, the algorithm factory module includes:
[0023] The algorithm development unit provides an online editor based on the security and compliance dataset output by the data security and compliance verification module. It supports users in developing custom algorithms according to the dataset type and outputting the initial algorithm code.
[0024] The algorithm verification unit, based on the initial algorithm code output by the algorithm development unit, selects the standardized dataset library in the data configuration module as the verification set to verify the algorithm and outputs the verified algorithm.
[0025] The training task configuration unit configures the training environment and training parameters based on the standardized dataset in the algorithm and data configuration module output by the algorithm verification unit, and outputs the training task configuration scheme.
[0026] The algorithm training unit, based on the training task configuration scheme of the training task configuration unit, executes the algorithm training process and outputs training process data in real time.
[0027] As a preferred embodiment of the brain-based autonomous decision-making system described in this invention, the cognitive inference calculation module includes:
[0028] The scenario configuration unit configures scenario parameters and outputs scenario-based scenario configuration based on the trained algorithm and real-time training process data output by the algorithm factory module, combined with actual decision-making needs.
[0029] The dynamic simulation unit performs autonomous simulations based on the scenario-based simulation configurations output by the simulation scenario configuration unit, and outputs initial simulation results and simulation process logs.
[0030] The simulation optimization unit compares the initial simulation results and simulation process logs output by the dynamic simulation unit with the preset thresholds. If the thresholds are not met, the algorithm parameters or simulation scenario parameters are adjusted in reverse, and the optimized simulation algorithm and simulation results are output.
[0031] The deduction result evaluation unit evaluates the deduction results based on the optimized deduction results output by the deduction optimization unit, combined with the original data characteristics in the data configuration module, from multiple dimensions, and outputs a qualified deduction model prototype.
[0032] As a preferred embodiment of the brain-based autonomous decision-making system described in this invention, the basic model management module includes:
[0033] The model encapsulation unit encapsulates the model framework and converts the format based on the prototype of the inference model output by the cognitive inference calculation module, and outputs a standardized model.
[0034] The model optimization unit optimizes the model based on the standardized model output by the model encapsulation unit and the newly added dataset in the data configuration module, using optimization algorithms to optimize the model and output the optimized model and optimization log.
[0035] The model version management unit establishes a model version repository based on the optimized model and optimization logs output by the model optimization unit, and outputs usable models with version identifiers.
[0036] The model deployment unit, based on the available models output by the model version management unit, configures online service parameters, deploys the models to service nodes, and outputs callable model service interfaces.
[0037] As a preferred embodiment of the brain-based autonomous decision-making system described in this invention, the decision output module includes:
[0038] The real-time data access unit, based on the data configuration module, accesses real-time scene data and outputs standardized real-time data.
[0039] The decision invocation unit, based on the standardized real-time data and basic model management module model service interface of the real-time data access unit, initiates decision requests and outputs model decision results;
[0040] The decision result verification unit, based on the decision results output by the decision call unit and the historical dataset in the data configuration module, compares the historical decision cases to verify the rationality and accuracy of the decision results, and outputs the verified decision results.
[0041] The decision output and feedback unit outputs the final decision conclusion based on the decision result output by the decision result verification unit, and simultaneously feeds back the decision result and verification information to the cognitive inference calculation module and the basic model management module.
[0042] Compared with existing technologies:
[0043] This invention precisely addresses the core pain points of existing decision-making systems. By constructing a unified dataset management system through a data configuration module, it achieves centralized control and efficient adaptation of various data resources, fundamentally solving the problem of fragmented data resources. Relying on the deep collaboration between the algorithm factory module and upstream and downstream modules, it creates a tight closed loop for algorithm development, verification, and training, allowing for the reuse of previous resources without repeated environment configuration, effectively connecting the algorithm development and training processes. Through the dynamic monitoring and reverse optimization mechanism of the cognitive inference calculation module, it captures key inference indicators in real time and adjusts strategies, completely changing the black-box operation mode of inference and solving the problem of lack of dynamic monitoring and feedback. Through the integrated control of the basic model management module and the feedback mechanism of the decision output module, it achieves seamless connection and reverse iteration between model optimization, deployment, and online services, overcoming the predicament of poor model management and service integration, and significantly improving overall system synergy and decision reliability. Attached Figure Description
[0044] Figure 1 This is a schematic diagram of the overall framework of the present invention;
[0045] Figure 2 This is a schematic diagram of the data configuration module framework of the present invention;
[0046] Figure 3 This is a schematic diagram of the data security and compliance verification module framework of the present invention;
[0047] Figure 4 This is a schematic diagram of the algorithm factory module framework of the present invention;
[0048] Figure 5 This is a schematic diagram of the cognitive inference calculation module framework of the present invention;
[0049] Figure 6 This is a schematic diagram of the basic model management module framework of the present invention;
[0050] Figure 7 This is a schematic diagram of the decision output module framework of the present invention. Detailed Implementation
[0051] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.
[0052] This invention provides a brain-based autonomous decision-making system; please refer to [link / reference]. Figure 1 ,include:
[0053] The data configuration module collects data from multiple sources, classifies, cleans, converts, and labels the data, builds a unified repository and indexes it, and outputs a standardized dataset library with indexes.
[0054] The data security and compliance verification module, based on the standardized dataset library of the data configuration module, configures scenario-based compliance rules, and outputs a security compliance dataset and traceability log after de-identification, security verification and logging.
[0055] The Algorithm Factory module, based on the security and compliance dataset output by the Data Security and Compliance Verification module, completes algorithm development, verification, training configuration and training, and outputs the trained algorithm and real-time training process data.
[0056] The cognitive inference calculation module, based on the trained algorithm and real-time training process data output by the algorithm factory module, configures the scenario, dynamically infers, optimizes and evaluates, and outputs a preliminary inference model that has passed the evaluation.
[0057] The basic model management module, based on the prototype of the inference model output by the cognitive inference calculation module, is encapsulated, optimized, version managed and deployed, and outputs a callable model service interface with version identifier.
[0058] The decision output module, based on the callable model service interface output by the basic model management module and the standardized dataset library of the data configuration module, outputs the final decision conclusion and provides feedback after access, decision invocation, and result verification.
[0059] Please see Figure 2 The data configuration module includes:
[0060] The dataset acquisition unit collects data from multiple sources, including structured data (such as tabular data) and unstructured data (such as text and images), and outputs the raw dataset.
[0061] The dataset classification unit classifies the raw data set output by the dataset collection unit according to data type (such as numerical, text, and image) and application scenario (such as risk control, transportation, and manufacturing), and outputs the classified raw data set.
[0062] The dataset processing unit performs data cleaning (duplicate removal, missing value completion), format conversion (standardizing to CSV, TXT, and other standard formats), and label annotation on the raw dataset output by the dataset classification unit, and outputs a standardized dataset.
[0063] The dataset storage and indexing unit builds a unified repository for custom datasets and pre-built datasets based on the standardized datasets output by the dataset processing unit. It also establishes multi-dimensional indexes such as dataset name, type, creation time, and application scenario to support fast querying and retrieval, thus outputting a standardized dataset library with indexes.
[0064] Please see Figure 3 The data security and compliance verification module includes:
[0065] The compliance rule configuration unit, based on the standardized dataset library of the data configuration module, combines the compliance requirements of different industries (such as finance and transportation) (such as the Data Security Law and the Personal Information Protection Law) to configure data usage permissions, sensitive data identification rules and compliance verification standards, and outputs a scenario-based compliance rule set;
[0066] The data desensitization unit, based on the rule set of the compliance rule configuration unit and the standardized dataset library output by the data configuration module, performs desensitization processing (such as masking, encryption, and anonymization) on data containing sensitive information (such as personal ID numbers and core enterprise data), and outputs the desensitized standardized dataset.
[0067] The security verification unit, based on the standardized dataset output by the data anonymization unit, verifies the encryption status of data transmission, access permission adaptability, and anonymization integrity. It also intercepts illegal data (such as data beyond the scope of permissions or sensitive data that has not been anonymized) and outputs a security compliance dataset and verification report.
[0068] The compliance log recording unit, based on the verification report of the security verification unit, records data processing information such as the data desensitization process, compliance verification results, and data access trajectory, forming an immutable compliance log and outputting log files for subsequent traceability.
[0069] Please see Figure 4 The algorithm factory module includes:
[0070] The algorithm development unit provides an online editor based on the secure and compliant dataset output by the data security and compliance verification module. It supports users in developing custom algorithms (such as regression algorithms, classification algorithms, and object detection algorithms) according to the dataset type and outputting the initial algorithm code.
[0071] The algorithm verification unit, based on the initial algorithm code output by the algorithm development unit, selects the standardized dataset library in the data configuration module as the verification set to verify the syntactic correctness and runtime stability of the algorithm, and outputs the algorithm that has passed the verification.
[0072] The training task configuration unit, based on the standardized dataset (specifying the training set and validation set) in the algorithm and data configuration module output by the algorithm validation unit, configures the training environment (such as CPU / GPU nodes, image type) and training parameters (such as number of iterations, learning rate), and outputs the training task configuration scheme.
[0073] The algorithm training unit executes the algorithm training process based on the training task configuration scheme of the training task configuration unit, and outputs training process data in real time (such as loss function value, accuracy, and resource usage).
[0074] Please see Figure 5 The cognitive inference calculation module includes:
[0075] The scenario configuration unit configures scenario parameters and outputs scenario-based scenario configurations based on the trained algorithm and real-time training process data output by the algorithm factory module, combined with actual decision-making needs (such as fraud detection in risk control scenarios and traffic flow prediction in traffic scenarios).
[0076] The dynamic simulation unit executes autonomous simulations based on the scenario-based simulation configurations output by the simulation scenario configuration unit, and monitors key indicators (such as simulation accuracy and running efficiency) in real time during the simulation process, outputting initial simulation results and simulation process logs.
[0077] The simulation optimization unit, based on the initial simulation results and simulation process logs output by the dynamic simulation unit, compares them with preset thresholds (such as accuracy thresholds and efficiency thresholds). If the thresholds are not met, it adjusts the algorithm parameters (by calling the algorithm development unit to optimize the parameters) or the simulation scenario parameters in reverse, and outputs the optimized simulation algorithm and simulation results.
[0078] The deduction result evaluation unit evaluates the deduction results based on the optimized deduction results output by the deduction optimization unit, combined with the original data characteristics in the data configuration module, from three dimensions: accuracy, efficiency, and adaptability, and outputs a preliminary deduction model that has passed the evaluation.
[0079] Please see Figure 6 The basic model management module includes:
[0080] The model encapsulation unit encapsulates the model framework (such as TensorFlow or PyTorch framework) and converts the format (such as ONNX or SavedModel format) based on the prototype of the inference model output by the cognitive inference calculation module, and outputs a standardized model.
[0081] The model optimization unit optimizes the model based on the standardized model output by the model encapsulation unit and the newly added dataset in the data configuration module, using optimization algorithms such as quantization, distillation, and pruning. It outputs the optimized model and optimization log (including comparison data of accuracy, size, and inference speed before and after optimization).
[0082] The model version management unit establishes a model version library based on the optimized model and optimization logs output by the model optimization unit. It records the parameters, optimization strategies, and applicable scenarios of each version, supports version query, rollback and update, and outputs available models with version identifiers.
[0083] The model deployment unit, based on the available models output by the model version management unit, configures online service parameters (such as HTTP / GRPC service type, node specifications, and canary release strategy), deploys the model to the service node, and outputs callable model service interfaces.
[0084] Please see Figure 7 The decision output module includes:
[0085] The real-time data access unit, based on the data configuration module, accesses real-time scenario data (such as real-time transaction data and real-time traffic flow data) and outputs standardized real-time data.
[0086] The decision invocation unit, based on the standardized real-time data and basic model management module model service interface of the real-time data access unit, initiates decision requests and outputs model decision results;
[0087] The decision result verification unit, based on the decision results output by the decision call unit and the historical dataset in the data configuration module, compares the historical decision cases to verify the rationality and accuracy of the decision results, and outputs the verified decision results.
[0088] The decision output and feedback unit outputs the final decision conclusion based on the decision result output by the decision result verification unit, using methods such as a visual interface and API push. At the same time, it feeds back the decision result and verification information to the cognitive inference calculation module and the basic model management module for subsequent model optimization and algorithm iteration.
[0089] Based on the aforementioned brain-driven autonomous decision-making system, this invention also provides a decision-making method, comprising the following steps:
[0090] S1: Data configuration stage: The dataset collection unit of the data configuration module collects raw data from multiple sources. After classification and processing, the storage and indexing unit builds a standardized dataset library.
[0091] S2: Data security and compliance verification phase: Call the standardized dataset output by S1, configure scenario-based compliance rules, generate a security and compliance dataset after de-identification and security verification, and record compliance logs at the same time;
[0092] S3: Algorithm Development and Training Phase: The Algorithm Factory module calls the safety and compliance dataset output by S2 to complete the algorithm development, verification, training task configuration and training, and outputs the trained algorithm and training process data.
[0093] S4: Cognitive Deduction Stage: The cognitive deduction calculation module combines the trained algorithm output from S3 with the scenario requirements to complete dynamic deduction, optimization and evaluation, and output a qualified model prototype.
[0094] S5: Model Management and Deployment Phase: The basic model management module encapsulates, optimizes, manages versions of, and deploys the prototype models output by S4, and outputs callable model service interfaces.
[0095] S6: Decision Output Stage: The decision output module accesses real-time data, calls the model service interface output by S5 to obtain the decision results, outputs the final decision conclusion after verification, and feeds back relevant information to the preceding modules to achieve iterative optimization.
[0096] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.
Claims
1. A brain-based autonomous decision-making system, characterized in that: include: The data configuration module collects data from multiple sources, classifies, cleans, converts, and labels the data, builds a unified repository and indexes it, and outputs a standardized dataset library with indexes. The data security and compliance verification module, based on the standardized dataset library of the data configuration module, configures scenario-based compliance rules, and outputs a security compliance dataset and traceability log after de-identification, security verification and logging. The Algorithm Factory module, based on the security and compliance dataset output by the Data Security and Compliance Verification module, completes algorithm development, verification, training configuration and training, and outputs the trained algorithm and real-time training process data. The cognitive inference calculation module, based on the trained algorithm and real-time training process data output by the algorithm factory module, configures the scenario, dynamically infers, optimizes and evaluates, and outputs a preliminary inference model that has passed the evaluation. The basic model management module, based on the prototype of the inference model output by the cognitive inference calculation module, is encapsulated, optimized, version managed and deployed, and outputs a callable model service interface with version identifier. The decision output module, based on the callable model service interface and standardized dataset library of the data configuration module output by the basic model management module, outputs the final decision conclusion and provides feedback after access, decision invocation, and result verification. The algorithm factory module includes: The algorithm development unit provides an online editor based on the security and compliance dataset output by the data security and compliance verification module. It supports users in developing custom algorithms according to the dataset type and outputting the initial algorithm code. The algorithm verification unit, based on the initial algorithm code output by the algorithm development unit, selects the standardized dataset library in the data configuration module as the verification set to verify the algorithm and outputs the verified algorithm. The training task configuration unit configures the training environment and training parameters based on the standardized dataset in the algorithm and data configuration module output by the algorithm verification unit, and outputs the training task configuration scheme. The algorithm training unit, based on the training task configuration scheme of the training task configuration unit, executes the algorithm training process and outputs training process data in real time.
2. The brain-based autonomous decision-making system according to claim 1, characterized in that, The data configuration module includes: The dataset acquisition unit collects data from multiple sources and outputs the raw dataset. The dataset classification unit classifies the raw data set output by the dataset collection unit according to data type and application scenario, and outputs the classified raw data set. The dataset processing unit performs data cleaning, format conversion, and label annotation on the raw dataset output by the dataset classification unit, and outputs a standardized dataset. The dataset storage and indexing unit builds a unified repository for custom datasets and pre-built datasets based on the standardized datasets output by the dataset processing unit. It also establishes multi-dimensional indexes to output a standardized dataset library with indexes.
3. The brain-based autonomous decision-making system according to claim 1, characterized in that, The data security and compliance verification module includes: The compliance rule configuration unit, based on the standardized dataset library output by the data configuration module, combines the compliance requirements of different industries to configure data usage permissions, sensitive data identification rules and compliance verification standards, and outputs a scenario-based compliance rule set; The data desensitization unit, based on the rule set of the compliance rule configuration unit and the standardized dataset library output by the data configuration module, performs desensitization processing on data containing sensitive information and outputs the desensitized standardized dataset. The security verification unit, based on the standardized dataset output by the data anonymization unit, verifies the encryption status of data transmission, access permission adaptability, and anonymization integrity, while also blocking illegal data and outputting a security compliance dataset and verification report. The compliance log recording unit records data processing information and outputs log files based on the verification report from the security verification unit.
4. The brain-based autonomous decision-making system according to claim 1, characterized in that, The cognitive deduction and calculation module includes: The scenario configuration unit configures scenario parameters and outputs scenario-based scenario configuration based on the trained algorithm and real-time training process data output by the algorithm factory module, combined with actual decision-making needs. The dynamic simulation unit performs autonomous simulations based on the scenario-based simulation configurations output by the simulation scenario configuration unit, and outputs initial simulation results and simulation process logs. The simulation optimization unit compares the initial simulation results and simulation process logs output by the dynamic simulation unit with the preset thresholds. If the thresholds are not met, the algorithm parameters or simulation scenario parameters are adjusted in reverse, and the optimized simulation algorithm and simulation results are output. The deduction result evaluation unit evaluates the deduction results based on the optimized deduction results output by the deduction optimization unit, combined with the original data characteristics in the data configuration module, from multiple dimensions, and outputs a qualified deduction model prototype.
5. The brain-based autonomous decision-making system according to claim 1, characterized in that, The basic model management module includes: The model encapsulation unit encapsulates the model framework and converts the format based on the prototype of the inference model output by the cognitive inference calculation module, and outputs a standardized model. The model optimization unit optimizes the model based on the standardized model output by the model encapsulation unit and the newly added dataset in the data configuration module, using optimization algorithms to optimize the model and output the optimized model and optimization log. The model version management unit establishes a model version repository based on the optimized model and optimization logs output by the model optimization unit, and outputs usable models with version identifiers. The model deployment unit, based on the available models output by the model version management unit, configures online service parameters, deploys the models to service nodes, and outputs callable model service interfaces.
6. The brain-based autonomous decision-making system according to claim 1, characterized in that, The decision output module includes: The real-time data access unit, based on the data configuration module, accesses real-time scene data and outputs standardized real-time data. The decision invocation unit, based on the standardized real-time data and basic model management module model service interface of the real-time data access unit, initiates decision requests and outputs model decision results; The decision result verification unit, based on the decision results output by the decision call unit and the historical dataset in the data configuration module, compares the historical decision cases to verify the rationality and accuracy of the decision results, and outputs the verified decision results. The decision output and feedback unit outputs the final decision conclusion based on the decision result output by the decision result verification unit, and simultaneously feeds back the decision result and verification information to the cognitive inference calculation module and the basic model management module.
