reverse-gol

Reverse-GoL

This is my Kaggle entry for the "Reverse Game-of-Life" competition that ended in May 2014.

The general approach is to use Genetic Algorithms to generate precursor board(s) that target the desired end-state.
As a sub-goal, I used this as an opportunity to create something self-contained in GoLang.

In general, however, the top of the leaderboard was dominated by people with the computing power (and smarts) to enumerate the precursor boards exhaustively.

Therefore, while my approach is interesting, it's main value is in the 'lessons learned' : And, of course, whatever useful code you can find in this project.

I also have to emphasise that this project was coded rather rapidly, particularly in the end-stages - so there are definitely a lot of rough edges (although some parts are quite refined - it's just really uneven...)

Presentation

There's a presentation about the GoLang / Kaggle experience at http://RedCatLabs.com/2014-10-14_Reverse-GoL/

Installation

git clone <ThisRepo>
cd <ThisRepo>
GOPATH=`pwd` go build reverse-gol.go speed_packed.go ga.go board-standard.go transitions.go db.go && ./reverse-gol

Installation of MySQL library :

GOPATH=`pwd` go get github.com/go-sql-driver/mysql

Then, (for definiteness, obviously you can change these defaults by editing db.go:48) create a database user 'reverse-gol' with password 'reverse-gol' with access rights to database 'reverse-gol'.

The MySQL table creation commands are at the beginning of db.go - and you'll have to run these manually yourself.

Running

To compile and run, use the following :

GOPATH=`pwd` go build reverse-gol.go speed_packed.go ga.go board-standard.go transitions.go db.go && ./reverse-gol

To see the different use-cases of this only-built-for-results code, do a ./reverse-gol --help, and then examine the source...

Usage:
  -cmd="": Required : {db|create|visualize|run|submit}
  -count=0: Number of ids to process
  -delta=0: Number of steps between start and end
  -id=0: Specific id to examine
  -seed=1: Random seed to use
  -training=false: Act on training set (default=false, i.e. test set)
  -type="": create:{fake_training_data|training_set_transitions|synthetic_transitions}, db:{test|insert_problems}, visualize:{data|ga}, submit:{kaggle|fakescore}