An improved algorithm model based on machine learning

In the last decades, Reinforcement Learning (RL) algorithm has attracted more and more attention, and become the research focus in the field of machine learning. This paper leads the typical RL algorithm, Q-learning algorithm, into computer game platform (Connect6), and proposes an improved method. We adjust reward parameter according to the shape of Connect6, and optimize the adjustment of evaluation function to achieve the global optimization. Moreover, the optimization of the reward makes the valueless units away from the evaluation, to reduce the interference of valueless units for optimal results and improve the convergence speed, thereby reducing the overall time of self-learning process.