Learning to control agents directly from high-dimensional game spaces is one of long-standing challenges of reinforcement learning. We will then cover our approach, results and any further direction that we could have potentially taken. ĭuring this paper we will talk more about the previous work done to to solve Atari games and what what research questions we had specifically for road crossing type of games. The most current technique at the moment are called Schema Networks which is a physics simulator that is capable of disentangling multiple causes of events and reasoning backward through causes to achieve goals. This was beneficial because the model converges slower and learns from multiple passes as well as having a better convergence. The way that DeepMind Technologies fixed this problem was using an experience replay mechanism, which allows the agent to learn from earlier memories can speed up learning and break undesirable temporal correlations. We can obviously see that Atari games specifically work by having states that are not independent but instead depend on heavily on previous states that could possibly have long chains of states. Some of the initial problems with using the Deep Learning modification were that Deep Learning models require there to be an abundance in trained label that are independent. Due to these Reinforcement Learning (RL) techniques, this model was able to outperform any previously used model in about 6 Atari games as well as outperform expert human players in 3 of those games. DeepMind Technologies created a variant of Q-Learning called Deep Q-learning that takes in a representation of the game as raw pixels and returns a value functions that is used to estimate the reward of future actions. Atari games were a region of interest for DeepMind Road Crossing games such as the Atari games, Freeway and Frostbite have many opportunities for machine learning research. Marlon Facey, Abraham Hussain, Jeffrey Pak, Kevin Tat, Michelle Zhao 1 Introduction Deep Q Learning for learning Crossy road and Frostbite games Reinforcement Learning for Road Crossing Games
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |