CE811: Game Artificial Intelligence Set: 16th October 2015 Due: Wednesday 2nd December 2015
Assignment I The first assignment for CE811 consists of writing and submitting an agent for the GVGAI framework. The agent must be submitted to the GVGAI website (www.gvgai.net), and it will take part on the private league organized for this module. The controller should be based in one of the techniques seen during the module (such as MCTS or RHEA), with some modifications to perform well in the GVGAI framework (for instance, pruning, targets, value functions, etc.). You must discuss the algorithms and these adjustments in the report, motivating them and comparing (in a rigorous and statistical way) their effects on the original algorithms. This assignment covers the following areas or skills: • Professional skills: Reading and writing scientific papers, and practising to present a technical talk. • Practical skills: Improving your programming skills, learning new algorithms and solving problems. • Soft skills: taking part in international communities and competitions. The assessment of this assignment is composed of the following parts: • Written report (20%). • Oral presentation (10%). The submission of the assignment must be done through Faser in a zip file, including the sources of your controller, the final report, (in PDF format) and the presentation slides.
Written Report A report must be written with the following suggested structure: • • • • • •
Introduction: initial introduction to the problem, motivation and outline of the rest of the paper. Literature Review: Study of what has been done before in general game (and video game) playing. Benchmark: Brief explanation of the problem. What’s GVGAI and what are its main challenges? Background: Explain here the algorithm that the submitted controller’s technique is based on. Techniques Implemented: Explanation of the agent submitted. Experimental Study: Describe the experimental study performed to obtain the final submitted agent (i.e., the procedure used to obtain the best combination of parameters, effect on the results of adding different functionality to the agent, etc.) • Overall conclusions and potential Future Work. • References.
The report must be written using one of the templates (Word or LaTeX) provided in the module page, and it should be between 15 and 20 pages long. The report will be assessed attending factors such as its contents, clarity, explanations, references, quality of the algorithm designed, etc. 1
Oral Presentation There will be a 10 minute presentation on week 10 (Friday 4th December) on CES Lab 5. This presentation should cover: • Description of the controllers implemented. • Results obtained with the different configurations. • Analysis of the development process: – What things worked? What things didn’t work? Did something surprise you? – Your state-value function. • Main problems and difficulties found. • Conclusions: main findings and lessons learned. • Future work. The presentation will be assessed according to its content, appropriate duration, clarity, entertainment, etc.
2
