Reinforcement learning model and model weight reduction and optimization method for esports strategy optimization

摘要

It relates to a method of optimizing an e-sports strategy for e-sports education, and a reinforcement learning model for strategy analysis and a method for lightening and optimizing the model are provided, and in this case, a reinforcement learning model algorithm for real-time game analysis is provided. In the Observation process of the reinforcement learning process, real-time observations are acquired from e-sports matches, the acquired observations are processed by the Deep RL Agent, and the input values of the Deep Neural Network are generated as a single batch. And the state value s(t) from the observation value, and the state value s(t) come out as a(t) action value through the inference process. ) to generate a value. The generated state, action, and reward values are stored in the Experience Buffer to minimize external memory access, and the Policy Network updates the weights of the Deep Neural Network and receives input values in Multi Batch In order to accelerate the operation, all environments are allocated and calculated in parallel in the simulator.