A reinforcement learning training method and a decision-making method based on reinforcement learning

Abstract

The present invention provides a reinforcement learning training method and a decision-making method based on reinforcement learning, wherein the reinforcement learning model training method includes the following steps: acquiring multiple sets of historical state data; inputting each set of historical state data into the reinforcement learning model to obtain Preliminary decision data; input each set of historical state data and the preliminary decision data into the pre-established Bayesian neural network model to obtain state change and reward value, and the state change value is the current state data and The difference of the next state data; updating the model parameters of the reinforcement learning model according to each set of historical state data and corresponding preliminary decision data, state change and reward value. By implementing the invention, the training sample size of the reinforcement learning model can be increased, the effect of reinforcement learning can be improved, and the accuracy of dynamic decision planning results can be improved.

Description

technical field [0001] The invention relates to the field of machine learning, in particular to a reinforcement learning training method and a decision-making method based on reinforcement learning. Background technique [0002] At present, reinforcement learning is an effective way of dynamic decision planning for multiple application scenarios, which focuses on how the subject should act in an environment to maximize cumulative rewards. Reinforcement learning application scenarios generally include transportation, finance, energy, business management and other fields, such as flight space management, through the use of reinforcement learning based on multiple status data of the flight (such as the number of remaining cabins, cabin sales, etc.) Control management. [0003] In related technologies, when performing reinforcement learning on the model, a large amount of state data is required to facilitate the training of the reinforcement learning model. In fact, the real st...

AI Technical Summary

Problems solved by technology

[0004] In view of this, the embodiment of the present invention provides a reinforcement learning training method and a reinforcement learning-based decision-making method to solve the defect of inaccurate dynamic decision planning results in the prior art

Images