10 results about "Learning rule" patented technology

The embodiments of the present disclosure disclose a training method of a semantic segmentation model, a semantic segmentation method and device of an image, wherein the method comprises: updating a first semi-supervised semantic segmentation model based on a first processing result of the first semi-supervised semantic segmentation model on first labeled training image data and a second processing result of the first semi-supervised semantic segmentation model on first unlabeled training image data, to obtain a second semi-supervised semantic segmentation model; determining at least one image to be labeled by using a preset active learning rule based on the second processing result; updating the first labeled training image data and the first unlabeled training image data based on the obtained each image to be labeled and the corresponding label; and training the second semi-supervised semantic segmentation model based on the updated second labeled training image data and the second unlabeled training image data to obtain a target semantic segmentation model. The embodiments of the present disclosure can effectively improve the distribution of labeled data, thereby effectively improving the performance of the semantic segmentation model.

The invention relates to the technical field of ocean platform pipeline design, in particular to an ocean platform pipeline laying method based on reinforcement learning, and the method comprises the following steps: S1, dispersing a pipeline laying region into a three-dimensional grid matrix according to the physical size of an actual cabin of an ocean platform, and marking a pipeline starting point, a pipeline ending point and an impassable region; s2, performing Q-Learning algorithm parameter initialization configuration, defining an action space adaptive to the linear movement characteristics of the ocean platform pipeline, constructing a Q value matrix adaptive to three-dimensional space coordinates and actions, and performing initialization; s3, entering a training round, and continuously updating the value evaluation matrix by dynamically adjusting a greedy criterion, a multi-dimensional reward mechanism and a time sequence difference learning rule; and S4, after the training is completed, starting from the starting point based on the converged Q value matrix, selecting an optimal action through a greedy to generate a final pipeline path, and improving the quality and search efficiency of pipeline laying on the ocean platform.

The application discloses an implementation method and system for self-evolution learning of an intelligent agent, and belongs to the technical field of intelligent agent reinforcement learning, large language models, memory and cognitive intelligence, and transfer learning; the method comprises the following steps: acquiring an environment state; inputting the environment state into a pre-trained strategy to output an optimal action and executing the action; after executing the action, a feedback signal is acquired; the feedback signal comprises an environment response and a task completion degree; and a new strategy is obtained by optimizing the strategy according to the feedback signal. The task performance of the application is continuously upgraded: the intelligent agent can be continuously optimized when facing long-range or repetitive tasks, and the intelligent agent will become more and more skilled in processing cross-platform complex tasks. The application reduces the research and application cost: the intelligent agent can reduce the dependence on manual work, can autonomously discover reinforcement learning rules, does not need to continuously manually annotate data, and relies on environment feedback for iterative optimization.

This invention discloses a frequency adaptive learning circuit and method with steady-state switching function, comprising a main circuit, a damping term circuit, a nonlinear term circuit, and a learning rule circuit. The input terminal of the main circuit serves as the input signal terminal of the frequency adaptive learning circuit, receiving a weak input characteristic signal and a high-frequency excitation auxiliary signal. The input terminals of the damping term circuit and the nonlinear term circuit are electrically connected to the output terminal of the main circuit, and their output terminals are also electrically connected to the input terminal of the main circuit. The input terminal of the learning rule circuit receives the weak input characteristic signal and the high-frequency excitation auxiliary signal, and its output terminal is electrically connected to the input terminal of the main circuit. The steady-state switching module in the learning rule circuit switches the frequency adaptive learning circuit between a monostable operating state and a bistable operating state. In the monostable operating state, the weak input characteristic signal is denoised, while in the bistable operating state, the weak input characteristic signal is amplified.

The application discloses a kind of cut cylinder and bypass heat supply multi-mode working condition self-adaptive switching decision system and method, belong to steam turbine control technical field, it includes obtaining the real-time operating parameter and external demand parameter of unit, obtains current working condition state vector;Current working condition state vector and the demand prediction parameter of future time period are carried out multidimensional rolling optimization calculation, and optimization decision instruction is generated;Using current working condition state vector index pre-stored working condition control mapping relationship set, continuous control instruction is calculated;Current working condition state vector, the execution feedback data of continuous control instruction and the execution effect data of optimization decision instruction are associated with evaluation, and rule update instruction is generated;Rule update instruction is executed, and working condition control mapping relationship set is revised.The application uses the strategy that prospective rolling optimization and closed loop self-learning rule revision are combined, and the economic, stable and intelligent adaptive control of mode switching of steam turbine under variable working condition can be realized.

The application relates to the field of computer word processing, in particular to a Chinese blind translation method, system and device combined with rules. The method comprises the following steps: obtaining Chinese character data to be translated; inputting the Chinese character data to be translated into a large language model to obtain Braille translation; the large language model comprises a word embedding layer, a learnable rule layer, an encoding layer and a decoding layer; the Chinese character data to be translated is input into the word embedding layer to generate a word vector, and is input into the learnable rule layer to generate a rule vector; the word vector and the rule vector are fused to obtain a constraint enhanced feature vector; the constraint enhanced feature vector is input into the encoding layer to be encoded, and then is decoded through the decoding layer to obtain the Braille translation. The application combines Braille conversion rules, can solve the problem of converting Braille for multi-sound Chinese characters, and has good application value.

The present application belongs to the technical field of temperature control of fuel cell combined heat and power system, and discloses a temperature control method and system of fuel cell combined heat and power, which comprises generating proportional, integral and differential state quantities, adjusting the weight of each state quantity on line based on learning rules, and determining a feedback control quantity; based on power steady-state mapping relationship and dynamic change characteristics, respectively calculating steady-state feedforward component and dynamic feedforward component, and determining feedforward control quantity according to the steady-state feedforward component and the dynamic feedforward component; according to the feedback control quantity and the feedforward control quantity, superimposing to obtain the secondary side water pump speed, generating the control instruction of the secondary side water pump speed based on the secondary side water pump speed, and completing the temperature control of the stack inlet according to the control instruction. The present application constructs a composite control structure composed of power feedforward compensation and single neuron PID feedback control in parallel, and has the characteristics of fast dynamic response, strong adaptive ability, simple implementation and strong robustness.

This application discloses a voltage reduction method and system for power distribution systems based on an expert knowledge base, belonging to the field of power distribution system voltage regulation technology. The method includes: S1: constructing a dynamically evolving expert knowledge base, which includes domain knowledge, experiential strategies, and learning rules generated through online learning; S2: generating an initial voltage reduction strategy based on the expert knowledge base and combined with real-time collected power distribution network and new load operation data; S3: exploring and optimizing the voltage reduction strategy under knowledge constraints using an online learning engine guided by the expert knowledge base, generating optimization rules and feeding them back to the expert knowledge base for updating; S4: executing voltage reduction actions based on the updated expert knowledge base through a multi-granularity hierarchical regulation architecture; S5: collecting the effect data after voltage reduction regulation, feeding it back to the expert knowledge base, triggering the dynamic evolution of the knowledge base, and returning to step S2 to form a closed-loop optimization.

The application discloses a kind of setting methods, devices, media and equipment of relay protection setting value, belong to setting value setting field, and the application is obtained by acquiring the multi-source data of main grid, distribution network and user side, and is preprocessed and topologically dynamic mapping using digital twin modeling model, and constructs the digital twin that reflects the operation state of three-level distribution network.Based on this, combined with the short-circuit current calculation model of multiple types of distributed energy, the total short-circuit current of the region is calculated, and the relay protection setting rule parameters are dynamically adjusted through the reinforcement learning rule optimization model. With the main grid setting value as the constraint, the distribution network and user side setting value are matched, and the preliminary setting value scheme is generated. The scheme is checked using the sensitivity dynamic checking model, and finally a reliable distribution network relay protection setting value is generated through the blockchain collaboration model, ensuring the reliability and credibility of the setting value scheme. The application effectively solves the problem that the existing technology cannot accurately and efficiently set the relay protection setting value of the distribution network.

The application discloses a kind of setting methods, devices, media and equipment of relay protection setting value, belong to setting value setting field, and the application is obtained by acquiring the multi-source data of main grid, distribution network and user side, and is preprocessed and topologically dynamic mapping using digital twin modeling model, and constructs the digital twin that reflects the operation state of three-level distribution network.Based on this, combined with the short-circuit current calculation model of multiple types of distributed energy, the total short-circuit current of the region is calculated, and the relay protection setting rule parameters are dynamically adjusted through the reinforcement learning rule optimization model. With the main grid setting value as the constraint, the distribution network and user side setting value are matched, and the preliminary setting value scheme is generated. The scheme is checked using the sensitivity dynamic checking model, and finally a reliable distribution network relay protection setting value is generated through the blockchain collaboration model, ensuring the reliability and credibility of the setting value scheme. The application effectively solves the problem that the existing technology cannot accurately and efficiently set the relay protection setting value of the distribution network.