Skip to content
Strategery game and player for Machine Learning course.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
experiments
maps
src
test
.gitattributes
.gitignore
README.md
build.xml
default.conf
merge.php
test.conf

README.md

What is this?

A simple game which looks a bit like a simplified version of Risk. It was implemented as the environment for building a player that learns through reinforcement learning. This project is an assignment for the Machine Learning course 2012.

Compiling

Run ant compile in the root directory. This should build your classes in build/classes.

If things behave weird, or methods cannot be found, try recompiling by first running ant clean and then again ant compile.

Running Experiments

Create a configuration file for each configuration you want to test. For example, if I want to test a different number of hidden units, create two files:

nn-1.conf:

[network]
hidden_units  i 	1

nn-10.conf:

[network]
hidden_units  i   10

Now I can run these two experiments by calling:

java -cp build/classes TestGame ./default.conf ./nn-*.conf

Note that I first pass default.conf as this contains all the other configuration options. Only the parameters named in the individual experiment configurations will be altered for that particular experiment.

After all the experiments have finished, a summary will be printed. Also, for each run of each experiment a file named config.conf-run-#.csv will be created, in which config.conf is replaced with the path of the configuration file for that experiment, and # is the number of the run.

You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.