Implementation and evaluation of a virtual computer game agent with neural networks

• Κύριος συγγραφέας: Κοντοθανάσης, Ιωάννης
• Άλλοι συγγραφείς: Kontothanasisis, Ioannis
• Γλώσσα: English
• Έκδοση: 2022
• Θέματα: Reinforcement learning; Neural networks; Atari breakout; Ενισχυτική μάθηση; Νευρωνικά δίκτυα

The challenges of applying reinforcement learning to modern AI applications are interesting, particularly in unknown environments in which there are delayed rewards. Classic arcade games have garnered considerable interest recently as a test bed for these kinds of algorithms. The purpose of this thesis is to create a model that can learn policies and optimal behaviors by interacting with the environment. The model is a convolutional neural network, trained with a variant of Q-learning, whose input is raw pixels and whose output is a value function estimating future rewards. This model is applied to the Atari game Breakout from the Arcade Learning Environment with no adjustment of the architecture or learning algorithm. The agent learns to play in an advanced level the Atari Breakout game.