19 results about "Online strategy" patented technology

The invention discloses a hybrid heterogeneous cloud workflow scheduling method based on reinforcement learning, and belongs to the technical field of cloud computing. The method comprises the following steps: aiming at a cloud workflow scheduling problem, by taking minimization of completion time as a target and taking cost and resources as constraints, a three-dimensional collaborative constraint model is constructed, and the cost constraints comprise server-free function budget and virtual machine budget; integrating the hyper-heuristic framework into a reinforcement learning algorithm; an improved reinforcement learning algorithm is adopted to solve the cloud workflow scheduling problem, optimal execution resources are selected, and an optimal scheduling scheme is obtained; and performing real-time scheduling according to the optimal scheduling scheme, and introducing a deviation feedback mechanism to monitor an execution error in real time. According to the method, the workflow scheduling problem is decomposed into a closed-loop optimization process of state perception and action decision, dynamic environment perception, multi-target tradeoff and online strategy optimization are deeply fused, the global search function is achieved, local optimization can be achieved, the algorithm complexity is low, and robustness is high.

The invention discloses an autonomous controllable environment-oriented power business system tuning method and device, and a medium. The method comprises the steps of obtaining a business load fluctuation rate and a resource health index of a power business system; according to the business load fluctuation rate and the resource health index, a target strategy model is determined, and the target strategy model comprises at least one of an offline strategy model, an online strategy model and a preset rule model; and obtaining target feature data of the power business system, and inputting the target feature data into the target strategy model to obtain a target resource scheduling strategy of the power business system. The problems that an existing power business system management mode is insufficient in static performance and flexibility and poor in uncertainty processing capacity are solved, and the scheduling strategy for optimizing power business system resources is achieved. When the power business system oriented to the autonomous controllable technology is in the initial stage of construction, an artificial intelligence method is applied to finally realize interpretable performance adjustment and optimization, and the performance index of the business system after credential adaptation transformation is improved.

The invention discloses an enterprise appeal intelligent perception and closed-loop processing system based on multi-modal AI, and relates to the technical field of intelligent government affairs, the system realizes accurate semantic understanding and deep intention recognition of multi-modal fusion, and the intelligent perception and closed-loop processing of enterprise appeals are realized through a cross-modal attention mechanism and a unified semantic representation model. According to the method, deep fusion and complementary analysis of multi-source heterogeneous data such as voices, texts, images and the like are realized, the limitation of a traditional single-mode or simple splicing mode is broken through, the accuracy and robustness of intention recognition are remarkably improved, and misjudgment is fundamentally reduced; the method comprises the following steps: constructing a dynamic self-adaptive routing mechanism based on Actor-Critic reinforcement learning, constructing a work order assignment problem into a sequence decision problem, driving a model to learn a dynamic fusion and weight distribution strategy of multiple decision factors through a reward mechanism, and adopting an online strategy iteration optimization mechanism to realize an optimal assignment decision, so as to improve the work order assignment efficiency. And the shunting accuracy and efficiency are obviously improved.

The invention relates to a visual language slow thinking reasoning-oriented semi-offline strategy reinforcement learning method, aims to solve the problem of insufficient reasoning ability of a current large-scale visual language model (LVLM) in a complex multi-modal task, and relates to an extensible semi-offline strategy reinforcement learning (SOPHIA) framework. Comprising the steps of constructing a semi-off strategy behavior model combining online strategy visual understanding and off-line strategy reasoning, designing a feedback and distribution mechanism of visual and reasoning rewards and designing an off-line strategy optimization method based on the visual and reasoning rewards. Compared with the prior art, the visual slow-thinking reasoning ability of the LVLM is systematically improved, meanwhile, the defects of an existing method in the aspects of visual understanding consistency and reasoning generalization ability are overcome, and the visual slow-thinking reasoning ability of the LVLM is improved.

The application relates to the technical field of power system automation, and particularly discloses an intelligent substation automation management system. The system comprises a digital twin synchronous management module, a multi-engine collaborative deduction module, an online strategy optimization and simulation module, and an instruction safety checking and feedback module. The system builds a digital twin synchronized with a physical station as a unified analysis environment; multiple professional analysis engines interacting through a shared cognitive state library are run in the twin, interwoven evaluation and joint deduction of equipment state, protection logic and network reliability are realized; based on the deduction result, online simulation iteration is used to search for a multi-step optimization control plan in the twin environment; and after safety checking of the plan instruction, the instruction is executed, and feedback data is collected to drive the closed-loop evolution of the model and the strategy. The application realizes multi-source data fusion, dynamic collaboration of cross-professional analysis and adaptive optimization of control strategies, and improves the integrity and autonomy of intelligent substation management.

The invention relates to the technical field of education informatization, and particularly discloses a dynamic feedback type resource management system for an education research platform. The system comprises a real-time data acquisition module, a dynamic feature extraction module, a strategy generation module, a feedback evaluation module and a resource scheduling execution module. By constructing a complete technical chain of real-time data acquisition, dynamic feature extraction, online strategy generation, multi-dimensional feedback evaluation and resource scheduling execution, closed-loop dynamic optimization of the educational research platform resource management process is realized. The system can sensitively capture short-time demand fluctuation caused by environmental changes, interest transfer or emergencies in research activities, and adjusts a resource allocation strategy on a minute-level time scale through an online learning mechanism, thereby effectively solving the problem of strategy lag caused by offline training and updating of a traditional system.

The present application relates to a kind of visual language slow thinking reasoning-oriented semi-offline strategy reinforcement learning method, to solve the problem of insufficient reasoning ability of current large-scale visual language model (LVLM) in complex multimodal task, the present application relates to scalable semi-offline strategy reinforcement learning (SOPHIA) framework, including the construction of semi-offline strategy behavior model combining online strategy visual understanding and offline strategy reasoning, the mechanism of visual and reasoning reward is designed back and distribution, and three parts of offline strategy optimization method based on visual and reasoning reward.Compared with the prior art, the visual slow thinking reasoning ability of LVLM is systematically improved, and the deficiencies of existing methods in visual understanding consistency and reasoning generalization ability are overcome, and the visual slow thinking reasoning ability of LVLM is improved.

The embodiment of the invention relates to the technical field of service release, and discloses a cloud service online method and device based on intelligent perception and automatic arrangement, equipment and a medium, and the method comprises the steps: obtaining a service dependence topological graph corresponding to a to-be-online service cluster; performing scheduling analysis processing on the service dependency topological graph based on an intelligent scheduling algorithm to generate a service online strategy; acquiring a historical published log, and performing similarity processing on the service online strategy and the historical published log by using an LSTM time sequence model to obtain an optimized online strategy for the service online strategy; and online the to-be-online service cluster based on the optimized online strategy. The service dependency topological graph is analyzed by using an intelligent scheduling algorithm, and an optimal online strategy conforming to the service dependency topological graph is automatically generated, so that a cascade fault caused by dependency disorder of manual setting is effectively avoided, the service online efficiency and success rate are remarkably improved, and the service online fault rate is reduced.

The application provides a dynamic transformation strategy-based mimicry scheduling system, comprising: a scheduling manager, which is used for sending online and offline commands to control online time and online strategy of left and right scheduling modules; a left scheduling module, which is used for judging whether to switch to the mimicry scheduling system for execution body scheduling according to the online and offline commands; a right scheduling module, which is heterogeneous with the left scheduling module in design, and is used for judging whether to switch to the mimicry scheduling system for execution body scheduling according to the online and offline commands; a data buffer, which is used for receiving and buffering abnormal execution body information sent by a decider according to online and offline notifications of the left and right scheduling modules sent by the scheduling manager, and forwarding the abnormal execution body information to the scheduling module in the online state; and an output selector, which is used for switching the scheduling module in the online state to be connected with the data buffer according to the online and offline notifications of the left and right scheduling modules sent by the scheduling manager.

This invention discloses an intelligent financial risk control system based on ABTest technology, belonging to the field of financial risk control technology. It includes a process configuration module, a strategy management module, an ABTest management module, a decision execution engine, a verification and analysis module, a machine learning model integration module, and a real-time monitoring and early warning module. The system allows for the configuration of risk control decision-making processes in a visual manner and supports the flexible integration of various strategy tools and machine learning models. Utilizing ABTest and a champion-challenger (Champion Challenger) mechanism, it conducts parallel testing and scientific evaluation of online strategies and challenger strategies. Based on statistical indicators and model performance indicators, it automatically analyzes and recommends the optimal strategy, achieving intelligent switching and iteration of strategies. This invention features high flexibility, scalability, and intelligence, improving the accuracy, efficiency, and system stability of risk control decisions. It is suitable for the risk management and business optimization needs of financial institutions in rapidly changing business environments.

The application relates to the technical field of cloud computing and cloud service management, in particular to a cloud service adaptive combination method based on a Decision Transformer and a decision support system; the method unifies heterogeneous service data into hierarchical service vectors through hierarchical tokenization and service representation construction; a multi-scale interaction feature extraction and semantic enhancement mechanism is constructed to capture the semantic and behavior dependence between cloud services at different granularities to generate enhanced service representation and compatibility matrix; the state, action and target are jointly embedded into a Decision Transformer model to generate a cloud service combination strategy sequence; through adaptive combination optimization and backtracking adjustment mechanism, the execution feedback and environmental changes are dynamically evaluated, and multi-instance knowledge migration and online strategy generalization are combined to realize cross-instance and cross-scene decision support. The application realizes continuous output of an optimal service combination scheme in a dynamic cloud environment.

The application discloses a power distribution optimization scheduling method and system suitable for a commercial complex microgrid. In the modeling, a modular layered structure is introduced, and the whole system is divided into a main power source scheduling layer, a renewable output prediction response layer and an energy storage flexible adjustment layer. In the solving, a collaborative mechanism of mathematical optimization + reinforcement learning is adopted, an MILP model is used for accurate optimization on a deterministic part, and a reinforcement learning algorithm is used for online strategy generation on an uncertain part, so that fast iterative update and feedback control of the scheduling result are realized. Meanwhile, through a boundary coupling coordination algorithm, synchronous update of state variables among the sub-models and global feasibility of the scheduling solution are guaranteed, so that the response speed of the whole scheduling system, the solution quality and the system robustness are improved.

The present invention relates to an autonomous driving decision-making method based on offline-online fusion reinforcement learning, which relates to the fields of autonomous driving and artificial intelligence, and includes two stages: offline training and online strategy expansion. Initialize the strategies and offline value networks of offline and online reinforcement learning. In the offline training stage, the data in the static data set are placed in an offline experience pool, static data are randomly sampled from the offline experience pool, and the offline strategy and offline value network are updated. In the online strategy expansion stage, the offline strategy obtained in the offline training stage is first frozen and combined with the online strategy into an extended strategy set, with a certain probability, strategies are selected from the extended strategy set to interact with the environment and the interaction data are placed in the online experience pool, static data are randomly sampled from the online experience pool, and the online strategy and online value network are updated. While retaining the security of offline reinforcement learning, the present invention allows decision-making agents to benefit from further online learning.

The application provides a reinforcement learning method suitable for inhibiting performance fluctuation, a reinforcement learning intelligent agent containing an online policy network and a target policy network queue is constructed, a main experience pool and an evaluation experience pool are combined, and stable evolution and efficient convergence of a policy are realized. In an environment interaction process, the online policy network and the optimal policy in the target policy network queue are alternately sampled and obtain environment interaction information, and the environment interaction information is stored in the main experience pool. In a preset evaluation period, a parallel multi-environment rolling evaluation mechanism is used to obtain policy evaluation rewards, and the target policy network queue is dynamically updated or a policy rollback is performed based on the evaluation results. A double evaluation network is introduced, and an adversarial evaluation training mechanism based on double Q mean value and small double Q random sampling is used to inhibit value function overestimation and policy over-conservatism. Interaction data generated in the evaluation stage is stored in the evaluation experience pool, and the interaction data is used in combination with main experience pool data in the training process, so that the stability, robustness and convergence performance of the reinforcement learning training are improved.

The application provides a digital twin virtual-real mapping and intelligent control scheduling platform for a family scene, relates to the technical field of intelligent control, and comprises a virtual-real dynamic mapping engine, a behavior causal inference engine, a virtual pre-execution scheduling mechanism and a self-evolution scheduling mechanism.The platform calculates a four-dimensional credibility field vector through sensor data, realizes malignant distortion filtering and state self-healing mapping, constructs a family behavior causal graph by using a dynamic Bayesian network and a physical information neural network, realizes short and long double-cycle forward probability risk and macro-environment evolution prediction, in a trusted execution environment, uses a space-time correlation conflict matrix and a multi-objective game model to perform sandbox simulation and pre-performance and trajectory audit interception on concurrent instructions, solves a Pareto optimal path, and finally realizes online strategy incremental reconstruction and rule automatic emergence evolution by combining user fine-tuning feedback and environment long-lag response time lag through a space-time decay reinforcement learning algorithm.

The invention relates to the technical field of cloud computing and cloud service management, in particular to a Decision Transform-based cloud service adaptive combination method and a decision support system. The method comprises the following steps: uniformly converting heterogeneous service data into hierarchical service vectors through hierarchical tokenization and service representation construction; constructing a multi-scale interactive feature extraction and semantic enhancement mechanism, and capturing semantic and behavior dependencies among cloud services on different granularities to generate an enhanced service representation and compatibility matrix; the state, the action and the target are jointly embedded into a Decision Transform model, and a cloud service combination strategy sequence is generated; the execution feedback and environment change are dynamically evaluated through an adaptive combination optimization and backtracking adjustment mechanism, and cross-instance and cross-scene decision support is realized in combination with multi-instance knowledge migration and online strategy generalization. According to the invention, the optimal service combination scheme is continuously output in the dynamic cloud environment.

The invention relates to the technical field of browser availability assistance in artificial intelligence, and discloses a webpage self-adaption method and system based on multi-modal emotion perception. According to the method, multi-source initial signals of a user are collected, synchronization and preprocessing are carried out, multi-modal fusion reasoning is carried out on a local edge model, and emotion vectors and confidence coefficients are obtained. And selecting a strategy packet S from a strategy library in combination with page context classification and an individualized threshold value, executing webpage rendering adjustment on DOM, CSS and ARIA attributes through browser content scripts, and dynamically changing information density, typesetting readability, dynamic effect and interaction complexity. User feedback is recorded, and online strategy updating with constraints is adopted locally; when the resources are limited, adaptive degradation is executed; under privacy and authority management, only intermediate features are stored, and original audios and videos are not sent out. According to the method, real-time closed loop and individualized self-adaption of emotion-strategy-page are realized, and the internet experience of special crowds with attention disorder, anxiety and the like is remarkably improved.

The application relates to the field of browser usability auxiliary technology in artificial intelligence, and discloses a webpage self-adaptive method and system based on multi-modal emotion perception. The method collects multi-source initial signals of a user, performs synchronization and pretreatment, carries out multi-modal fusion reasoning in a local edge model, obtains an emotion vector and a confidence degree, selects a strategy package S from a strategy library in combination with page context classification and individualized threshold value, executes webpage rendering adjustment on DOM, CSS and ARIA attributes through a browser content script, dynamically changes information density, readability of layout, dynamic effect and interaction complexity, records user feedback, and adopts online strategy updating with constraints on the local side; adaptive degradation is executed when resources are limited; only intermediate features are saved under privacy and permission management, and original audio and video are not externally transmitted. The application realizes real-time closed loop and individualized adaptation of emotion-strategy-page, and significantly improves the online experience of special groups such as people with attention disorders and people with anxiety.