PS Natural Computation:
GOJen - Evolution of ANN Input Features

Description

This project was created in the context of the course "Natural Computation" held by Helmut Mayer.

The goal of this project is the evolution of input features for ANN (artificial neural network) Go players. As the board size grows, so does the amount of information processed by ANN input features. To reduce workload, the number of input features can be reduced. This project analyzes the impact of input features on the ANN's fitness, to assess which can be excluded.

Group Members

• Siegmar Alber
• Axel Baumgartner
• Luca Debiasi
• Alexander Zrinyi

Milestones

1. 25.04.2009

   Project Kickoff

2. 01.04.2009

   Getting familiar with Go-Gameplay and GOjen.

3. 15.04.2009:

   Understanding the working concepts of GOjen (source-code).

4. 22.04.2009

   Understanding how to evolve and evaluate input features.

5. 06.05.2009

   Begin coding.

6. 13.05.2009

   Testing

7. 20.05.2009

   Finalizing project and report

Presentations

1. 18.03.2009: Introduction
2. 01.04.2009: Problem Analysis
3. 06.05.2009: Implementation

Work Document

Report (work in progress)