417 results about "Augmented learning" patented technology

The invention discloses a dynamic tool selection and optimization system and method oriented to large model external tool calling. The method comprises the steps that data input by a user is obtained, and standardization processing and feature extraction are carried out; performing multi-dimensional task analysis based on the extracted features to generate a task analysis result; constructing a tool feature space, wherein the tool feature space comprises functional features, performance indexes and resource requirements; performing tool selection and combination optimization by adopting a dynamic self-adaptive matching engine; executing intelligent call link optimization, including parallel scheduling and resource allocation; the calling process is monitored in real time, and multi-level anomaly detection and early warning are carried out; generating a multi-objective optimization strategy based on the monitoring data; and continuously optimizing and updating the calling strategy through a reinforcement learning mechanism. According to the method, the tool selection accuracy can be improved, the calling efficiency is optimized, and the system reliability is enhanced.

The invention discloses an equipment state sensing system based on peak shifting and danger situation sensing, which relates to the technical field of electric digital data processing and comprises a data acquisition unit, a model training and optimizing unit, a situation sensing unit, a peak shifting regulation and control unit and a system interface and user interaction unit. According to the method, the device state data feature similarity is dynamically calculated by using the Gaussian kernel function, the similarity matrix is adjusted in combination with the gradient update item, dynamic weighting is performed in feature extraction, and redundant features are reduced by adopting the adaptive discarding probability, so that the feature extraction efficiency of the model in a high-dimensional feature space is improved; the problem that the dimension reduction effect is affected by too many redundant features of the equipment state data is solved. Equipment state features are extracted layer by layer through a deep forest algorithm of a multi-layer random forest structure, a decision tree is optimized in combination with a dynamic adjustment weight mechanism and a reinforcement learning strategy, and the classification capability of the model for equipment state changes and abnormal states is enhanced.

The invention relates to the technical field of AI computing, in particular to a knowledge reinforcement learning method and system based on multiple agents. Modifying parameters of the large language model by using a supervised fine-tuning training method for the pre-trained large language model; constructing a knowledge base of a large language model, and vectorizing, storing and retrieving texts in various file formats; a multi-agent workflow optimization information processing and decision-making process is developed, a real-time problem is decomposed, a logic chain for solving the problem is constructed, and internet texts are obtained, filtered and summarized. Compared with the prior art, the method has the advantages that the knowledge quantity of the pre-trained large language model in the specific field is increased, the knowledge base is high in flexibility, the obtained text is high in real-time performance, certain interpretability is achieved and the like, and then the ability of the large language model to solve problems in the specific field is improved.

The invention relates to the field of converter station equipment parameter analysis, and provides a converter station equipment parameter analysis method and system based on multi-parameter fusion monitoring, and the method comprises the following steps: collecting multi-parameter characteristic data of converter station equipment; performing streaming processing and abnormal mode identification on the multi-parameter characteristic data to obtain equipment abnormity early warning information of the converter station equipment; constructing a time-varying graph neural network model based on the multi-parameter feature data, obtaining a space-time dependency relationship among parameters through a causal reasoning method, and dynamically adjusting the feature weight of the time-varying graph neural network model in combination with a reinforcement learning algorithm to obtain a running state evaluation model; and inputting the real-time state feature sequence into the operation state evaluation model to obtain an equipment operation state evaluation result, and visually displaying the equipment operation state evaluation result and the abnormal early warning information to obtain a comprehensive operation state evaluation report of the converter station equipment. According to the invention, real-time monitoring and evaluation of the operation state of the converter station equipment are realized.

In some implementations, a device may determine a set of features for the machine learning model. The device may determine, using off-policy evaluation, a performance level of the machine learning model associated with the set of features. The device may determine, based on the performance level, a ranking of the set of features. The device may generate, based on the ranking, one or more subsets of features from the set of features. The device may determine, using the off-policy evaluation, performance levels of the machine learning model for respective subsets of features from the one or more subsets of features. The device may select, based on the performance levels, a subset of features from the one or more subsets of features. The device may obtain, via the machine learning model, one or more recommendations that are based on input data, wherein the machine learning model uses the subset of features.

An example method is provided for training a machine-learned student sequence processing model, the method comprising: obtaining a respective input; obtaining, from the student machine-learned sequence processing model, a respective output corresponding to the respective input; generating a multiscale refinement objective configured to jointly distill knowledge from a teacher machine-learned sequence processing model and reinforce preferred behavior of the student machine-learned sequence processing model, wherein the multiscale refinement objective comprises: a first component based on a divergence metric characterizing, for the respective input, a comparison of a plurality of predictions of the student machine-learned sequence processing model to a plurality of predictions of the teacher machine-learned sequence processing model; and a second component based on a reinforcement learning signal associated with the respective output; and updating the machine-learned student sequence processing model based on the multiscale refinement objective.

The embodiment of the invention provides an intelligent learning path recommendation method and system based on a dynamic state. The method is applied to the technical field of intelligent learning recommendation, and comprises the following steps: calculating a skill adaptation weight, a load weight and an emergency degree weight in real time based on a dynamic state of a target employee, including a current skill improvement condition, a workload and a task emergency degree; combining the skill matching degree, the learning strength and the task association degree of each learning unit in the candidate learning path, and performing comprehensive evaluation by using a multi-weight scoring function; and sorting the paths according to the comprehensive score, generating a personalized recommended learning path set and sending the personalized recommended learning path set to an employee terminal, thereby realizing intelligent learning path recommendation with dynamic adaptation and accurate matching. According to the scheme, personalized learning path pushing aiming at actual post requirements and working states of the employees can be realized, correlation, urgency and acceptability of learning contents are improved, learning efficiency and task adaptability are remarkably enhanced, and the employees are helped to quickly compete with post targets.

Disclosed is a method for enhancement of a low-light image based on reinforcement learning and aesthetic evaluation. The method include: generating images of non-normal luminance under different lighting scenes, and constructing a training dataset for a reinforcement learning system based on the images; initializing the training dataset, a policy network, and a value network; updating, based on a no-reference reward score and an aesthetic assessment reward score, the policy network and the value network; completing model training and outputting an enhanced image result. By expanding the action space range defined in reinforcement learning, the enhancement operations for the input low-light image gain a greater dynamic range, offering higher flexibility for real-world scenarios and better meeting low-light image enhancement needs. Additionally, by incorporating the aesthetic quality assessment scores as part of the loss function, the enhanced image achieves better visual effects and higher subjective user evaluation scores.

The invention discloses a traffic signal and direction-variable lane cooperative control method based on multi-agent learning, and the method comprises the steps: making a decision through the iteration of a traffic signal control module and a lane guiding control module, and carrying out the operation at the same time to control the traffic; the lane guiding control module adopts a dual depth Q network algorithm to enhance a learning model, an intelligent agent selects actions based on traffic flow and traffic signal states to maximize the lane utilization rate, and an award is calculated based on lane queuing length difference and the maximum direction demand traffic flow ratio; the traffic signal control module agent selects an action based on the observed lane density and the current lane state and obtains an award according to the traffic flow satisfaction of the selected phase. According to the method, the lane utilization rate is improved by controlling lane guidance, and the dynamically changing traffic flow is controlled in space and time dimensions by using a reinforcement learning framework; higher stability and efficiency are shown, and the traffic capacity of the intersection is improved.

The invention provides a power grid infrastructure project material collaborative management method, which comprises the following steps: analyzing historical material use, power grid construction progress and external environment data through a reinforcement learning algorithm, dynamically adjusting model parameters, and predicting material demands; a low-power-consumption wide area network and an edge computing technology are used for monitoring and analyzing material states in a warehouse and a transportation process in real time, and a result is transmitted to the central management system; based on the prediction data, an optimized material purchasing plan is automatically generated, and the purchasing strategy is dynamically adjusted by considering the arrival time, the inventory state and the supplier capability; using an improved ant colony algorithm to optimize the transportation route and time of the materials from the warehouse to the construction site; using an intelligent sensor to record the material use condition, and carrying out preliminary data processing and analysis through edge calculation; and an information management platform is established to realize information sharing and cooperative work. According to the method, the intelligence and efficiency of material management are improved by integrating multiple technologies.

The invention discloses a frequency switching method and device of an embedded power terminal. The method comprises the steps that frequency switching is triggered by predicting frequency deviation and comparing the predicted frequency deviation with a preset frequency synchronization threshold value, and the transmitting power and the modulation mode in the switching process are optimized through a fuzzy control algorithm so as to suppress transient interference. During switching execution, the node synchronization state is monitored in real time, and an interruption mechanism is triggered when abnormity occurs. And after the switching is completed, calibrating the frequency parameter of each node, and updating the frequency drift compensation model. And dynamically optimizing the compensation model and the switching strategy by utilizing historical data and adopting a reinforcement learning algorithm, and synchronously updating the compensation model and the switching strategy to each node. By means of predictive triggering, intelligent parameter optimization, real-time monitoring and the like, the problems of transient interference, exception handling, model optimization and the like in the power terminal node frequency synchronization process are effectively solved. According to the method, the reliability and stability of power terminal node frequency synchronization are remarkably improved, and a powerful guarantee is provided for safe and stable operation of a power system.

A profit-aware Offloading Framework towards Prediction-assisted MEC Network Slicing is provided. Towards MEC network slicing, formulate the optimization problem of maximizing long-term ESP profits and decouple it into the sub-problems of Edge network Slicing (EnS) and Computation offloading Access (CoA). For the slicing sub-problem, use a gated recurrent neural network (GRNN) to accurately predict user requests in different regions, and then use the optimal partitioning of network slices with the predicted requests and expected demands; For the offloading sub-problem, incorporating results from slice partitioning, use an improved deep reinforcement learning with twin ritic-networks and delay mechanism, solving the Q-value overestimation and high variance for approximating the optimal offloading and resource allocation.

The invention provides a financial risk real-time management method and system based on multi-source data and a block chain, and the method comprises the steps: carrying out the real-time collection of library table data, webpage data and log data, endowing different data sources with dynamic weights according to the importance and real-time performance of the data sources, and adjusting the weight values through an adaptive algorithm; performing right confirmation operation on each piece of data by using a block chain technology, and endowing the data with a block chain certificate; establishing a risk assessment model, and continuously optimizing and dynamically adjusting the risk assessment model by combining a reinforcement learning algorithm with a traditional machine learning algorithm; financial risks are monitored in real time through a built-in multi-dimensional early warning index system; and in combination with an AIGC technology, carrying out deep analysis on a risk prediction result, and automatically generating a detailed risk analysis report. According to the invention, through the dynamically adjusted and optimized risk assessment model, the real-time and accurate financial risk management and prediction capability is provided, and the security and transparency of data are ensured.

The invention discloses a weak supervision video anomaly detection method based on prompt learning knowledge enhancement, and belongs to the technical field of video intelligent analysis. A video side gives a section of abnormal scene video, video sequence features and audio sequence features are obtained through a feature extraction network, then a trained and complete feature aggregation network is input to carry out multi-modal feature aggregation, an abnormal score is obtained through a score prediction network, and text representation is carried out based on prompt learning. A prompt template is constructed for abnormal video tags through a knowledge graph, semantic expansion is performed on normal tags through a plurality of learnable parameters, cross-modal alignment is performed on the normal tags and a video side, so that features of the video side are close to different normal semantics, knowledge enhancement is performed by introducing external information, positive abnormal boundaries of the video are learned, and the detection performance is improved. And finally, multi-task joint optimization is carried out through different loss functions, and abnormal video clip positioning is carried out.

The invention aims to provide a hotel price prediction system combining reinforcement learning and a long short-term memory network, and relates to the technical field of neural networks, and the system comprises a data acquisition multi-dimensional time sequence feature extraction unit which is used for acquiring historical operation data of a hotel, carrying out the multi-scale decomposition of a multi-dimensional time sequence through wavelet transform, and obtaining a multi-dimensional time sequence feature extraction unit; obtaining a wavelet coefficient as a multi-dimensional time sequence feature; the long and short-term memory network processing unit is used for inputting the multi-dimensional time sequence features into a long and short-term memory network as input and outputting a hidden state; and the price distribution prediction unit is used for using the price prediction model according to the hidden state, generating predicted price distribution, defining a Markov decision process, using the predicted price distribution as a reinforcement learning model, and generating a final price prediction result according to the predicted price distribution. According to the invention, the precision of price prediction is improved, and the coping capability of the system for market fluctuation is greatly enhanced.

The invention provides a low-power wide area network integrated sensing and communication method based on channel sensing and reinforcement learning, and the method comprises the steps: receiving an uplink signal transmitted by a terminal node through a gateway receiver, demodulating the uplink signal through employing a soft demodulation algorithm, and obtaining a bit log-likelihood ratio; the deep reinforcement learning model completes channel reconstruction of the corresponding link during communication according to the bit log-likelihood ratio; and calculating the current optimal network configuration according to the reconstructed channel so as to distribute the current optimal network configuration to each terminal node during next downlink transmission. According to the method, the key environment information is mined by using the deep reinforcement learning model, so that the gateway can allocate optimal network configuration to different terminal nodes according to the current network environment condition, and the decoding capability of the gateway is improved.

The invention relates to the technical field of air conditioner control, and discloses an air conditioner load data control method and system.The method comprises the steps that firstly, meteorological parameters and load parameters of an air conditioner are collected, and data cleaning, timestamp definition and standardization processing are conducted; then, defining a feature vector according to the processed data, generating a feature data set, and constructing a load prediction model by using a neural network algorithm; and pre-training the model by using the source domain data set through a transfer learning strategy, and enhancing the target domain data to generate a new sample. Next, parameters of the pre-training model are migrated to the new task model, and partial network layer parameters are frozen to form an initialized reinforcement learning model; and finally, training unfrozen layer parameters by using the target domain data set, and finely adjusting the model to realize real-time and accurate optimal setting of the air conditioner temperature, thereby achieving the purpose of load optimization control. Through combination of transfer learning and reinforcement learning, the generalization ability, robustness and adaptability of the model are effectively improved.

A learning data set enhancing method comprises the steps of: determining a first initial learning data set comprising a first learning-purpose natural language query, database information regarding the database, and a first learning-purpose SQL query corresponding to the first learning-purpose natural language query; generating first novel database information regarding a first novel database having a different schema from the database by applying a first schema modification operation to the database; generating a first novel learning-purpose SQL query regarding the first novel database by applying a first SQL query synchronization operation corresponding to the first schema modification operation to the first learning-purpose SQL query; and determining a first novel learning data set comprising the first learning-purpose natural language query, the first novel database information, and the first novel learning-purpose SQL query.

The present invention relates to a method and system for detecting an anomaly of an electrical submersible pump, ESP. Specifically, the proposed invention provides: physics-based anomaly assessment, automatic data imputation using virtual sensors network (203), automatic continual learning based on automatic model drift detection, automatic failure prevention based on reinforcement learning, remaining useful life (RUL) estimation, and cohort analysis.

A method of distributed training of a machine learning model is provided. The method includes inputting a first set of observations and a first set of actions for a primary agent to generate a first Q function for the primary agent, inputting a second set of observations and a second set of actions for a set of secondary agents to generate a set of Q functions for the set of secondary agents, and generating a Qtot function from the first Q function and the set of Q functions by a mixing network for the primary agent, the Qtot function to generate actions or predictions to configure a first node to operate in a telecommunication network.

An energy storage system and a thermal management method therefor are provided. The method includes: loading a multi-agent reinforcement learning model pre-trained and optimized through a simulation environment; receiving state observation data at a current time instant; inputting the state observation data into the multi-agent reinforcement learning model for reinforcement learning and reasoning to output multi-control action information; and generating a control action information instruction based on the multi-control action information, transmitting the control action information instruction to a cooling system in the energy storage system, and performing by the cooling system, thermal management on the energy storage system in response to the control action information instruction.

The invention relates to the technical field of reinforcement learning intelligent scheduling, in particular to an intelligent scheduling management method for prefabricated part production resources based on reinforcement learning. The specific implementation process comprises the steps of collecting order demands, material distribution and production pedestals in real time, and mapping the order demands, the material distribution and the production pedestals into production mold load tensors; when a scheduling request is triggered, inputting the production modulo tensor into a resource arrangement network, and performing search and reasoning by using a scheduling strategy based on a multi-head attention mechanism to generate a resource scheduling matching instruction; calculating a state difference tensor, and outputting an efficiency reward signal in combination with a delivery constraint and a cost constraint; and packaging the production modulo tensor, the resource scheduling matching instruction and the reward signal into semantic interaction experience, storing the semantic interaction experience into a scheduling experience playback pool for gradient modulation, and iteratively optimizing a scheduling strategy. According to the method, learning can be carried out from a large amount of historical data by utilizing reinforcement learning, iterative optimization of the scheduling strategy is realized, the calculation time consumption of scheduling instruction generation is reduced, and the flexibility of production scheduling is improved.

The invention discloses an image recognition system for artificial intelligence based on reinforcement learning, which relates to the technical field of image recognition and comprises an image input module, a preprocessing module, a feature extraction module, a reinforcement learning module, an adaptability adjustment module, a recognition output module, a self-optimization module and a performance monitoring module. The system has the advantages that image data of various devices can be compatible through the image input module, images with different resolutions are subjected to adaptive processing, the image quality is ensured, the feature extraction module can extract different level features of the images, the granularity and the range of feature extraction are accurately controlled by adjusting the number and the size of convolution kernels, and the image quality is improved. And the enhanced learning module performs learning and decision making by using a Q-learning algorithm, continuously optimizes an identification strategy and interacts with the adaptive adjustment module and the self-optimization module, and the functions act together, so that the image identification accuracy is powerfully improved.

The invention relates to an unmanned aerial vehicle cluster formation decision making and hunting method based on reinforcement learning, and belongs to the technical field of multi-agent formation control and decision making. The expected position corresponding to each agent and the control instruction of the dynamic model of the multi-agent are obtained, the central node does not need to quickly realize the expected formation configuration given by the decision-making system in a distributed mode, and the formation control problem of the multi-agent is solved; secondly, the formation of the multiple agents can be visually manipulated to adapt to dynamically changing tasks, and the formation manipulation problem of the multiple agents is solved; according to an actual confrontation task, learning and evolutionary online autonomous decision-making of the multi-agent system is realized, and the formation autonomous decision-making problem of the multi-agent system is solved; and finally, the interpretability and safety of learning decision are enhanced by utilizing a control model mechanism, and the interpretability and safety problems in the formation strategy learning process are considered.

The present disclosure relates to methods and systems that preserve privacy in a secure multi-party computation (MPC) framework in multi-agent reinforcement learning (MARL). The methods and systems use a secure MPC framework that allows for direct computation on encrypted data and enables parties to learn from others while keeping their own information private. The methods and systems provide a learning mechanism that carries out floating point operations in a privacy-preserving manner.

A waypoint, navigation controller and corresponding controlling methods are described, where the controller functions as a multiple input-multiple output, e.g., nonlinear angular velocity and linear speed controller for a land vessel such as a skid-steer vehicle. The controller and the controlling methods may be based on a fuzzy logic controller (alternatively referred to as “fuzzy controller”). The membership functions of the fuzzy controller may employ a trapezoidal structure with a symmetric rule-base. In addition, a Hierarchical Rule-Base Reduction (HRBR) is incorporated into the controller so as to select only the rules most influential on state errors by selecting inputs / outputs, determining the most globally influential inputs, and generating a hierarchy relating inputs via a Fuzzy Relations Control Strategy (FRCS). This disclosure is further directed to a fuzzy logic controller with Hierarchical Rule-Base Reduction (HRBR) and implemented as neural network and training of such a fuzzy logic controller via reinforcement learning based on an ANFIS actor.

The invention relates to the technical field of anesthesia management, in particular to an anesthesia medicine excess brain function influence assessment system applying artificial intelligence, which comprises a data acquisition module, a data preprocessing module, a model training module, a real-time assessment module, a risk early warning module, a treatment suggestion module, a system interaction module and a remote monitoring module. Through the combination of a convolutional neural network and a long-short-term memory network, spatial-temporal characteristics of brain wave and functional magnetic resonance imaging data are extracted, a brain function thermodynamic diagram is generated in real time, and the influence of anesthetic drugs is analyzed; the system is combined with a reinforcement learning algorithm to dynamically adjust the anesthetic dosage, a risk early warning module dynamically adjusts a threshold value through physiological parameters and tolerance of a patient, an expert team can provide medical assistance in real time through a remote monitoring module, and brain function safety of the patient in the operation or anesthesia process is ensured; according to the invention, the safety, accuracy and individuation level of anesthesia management are effectively improved.

A method for training a vehicular occupant monitoring system includes disposing a first camera at a test vehicle that captures first image data representative of electromagnetic radiation having a first range of wavelengths and a second camera that captures second image data representative of electromagnetic radiation having a second range of wavelengths. The method includes disposing markers at an occupant seated at the seat of the test vehicle that, based on the first range and the second range of wavelengths, are (i) represented in the second image data and (ii) not represented in the first image data. The method includes annotating, using one or more frames of captured second image data comprising the imaged markers, locations of the markers at the occupant within one or more frames of captured first image data and training the vehicular occupant monitoring system using the annotated frames.

The invention discloses a code generation model training method and device, electronic equipment and a storage medium. A test case is obtained, and tree search is executed based on the test case; obtaining an evaluation value of each node in tree search, and performing backward transmission update according to the evaluation value of each node so as to update a state action value and an access count of each node; obtaining a correct search path in the tree search according to the state action value and the access count, and storing the correct search path in a model fine tuning database; and training a code generation model based on the model fine tuning database. Tree search is combined with a process reward model, pertinence and effectiveness of a search process are ensured, node evaluation and a backward transmission update mechanism enable the model to continuously optimize a state action value, and learning efficiency is enhanced. A code generation model is trained by a high-quality sample generated through reward guide tree search, and the capacity of the model for generating complex codes is improved.

The invention provides a large model parameter optimization and self-adaptive adjustment method based on reinforcement learning, and relates to the technical field of optimization and self-adaptive adjustment of large model parameters based on reinforcement learning in machine learning, and the method comprises an overall architecture fusing a large model main body, a reinforcement learning module, an environment perception module and a reward feedback mechanism. The method is used for realizing dynamic optimization and adaptive adjustment of large model parameters, and can rapidly increase the learning rate according to environment feedback in the initial stage of large model training, so that the model parameters rapidly approach to the direction of an optimal solution, and the early-stage exploration time of training is greatly shortened. In the later stage of training, the learning rate can be accurately reduced, model oscillation is avoided, and it is ensured that the model is stably converged to a globally optimal solution. According to the dynamic adjustment mechanism, the number of iterations of training is effectively reduced, the training efficiency is greatly improved, a large number of computing resources and time cost can be saved, and for example, in large-scale image classification model training, the training time can be shortened by more than 30%.