View Proposal

---

Proposer

Patricia Vargas

Title

Learning to play games from experts

Goal

Video-game playing by learning from expert data

Description

In this project, we will assess the ability of classification models to create an intelligent agent for video-game playing by learning from expert data. The idea consists of first capturing a sample of play sessions from expert players to create a training data set. Next, we will apply different Machine Learning models and a Symbolic Classification model to create an intelligent agent that mimics the actions of the expert player and evaluate the extrapolation abilities for later stages. We will also evaluate different approaches that help to improve the extrapolation abilities of the model and assess the performance of the agent by how far they can play the game. Additionally, we will debug the symbolic models to understand the agent behaviour and improve their performance.

Resources

Background

Python programming, Artificial Intelligence (experience with Artificial Neural Networks and Symbolic Regression desired)

Url

Difficulty Level

Challenging

Ethical Approval

None

Number Of Students

1

Supervisor

Patricia Vargas

Keywords

Degrees

Master of Science in Robotics