Abstract:
This project is intended to provide a deep neural network model using a variety of machine learning techniques, including computer vision, reinforcement tree learning, and case-based reasoning. We are implementing these techniques to form a novel approach for training AI-bot on its gaming environment. Bot will use computational methods to “learn” information directly from data without relying on a predetermined equation as a model. The algorithms adaptively improve their performance as the number of samples available for learning increases.
The product is a project that will use “Keras library” of neural network model to train model using different reinforcement techniques. Objects will be recognized through computer vision. Game will be developed using gaming development toolkit. Training API’s will train the model by using GPU computational power. Bot will be added to the game.
