Project 4: Machine learning (Pente webapp)
Joshua Thomas & Shaun Meyer

TODO: 
  - Neural Network
  - Cleanup of code (removing old WLM stuff)

-=CGI Server=-

session.hpp: 
  - This library provides session functions to easily
    load/edit/describe the game sessions.
  
session.cpp: 
  - This is a simple test program which demonstrates functions from
    session.hpp.

wlm.hpp 
  - This header file describes the State and Weight objects as
    well as the Vhat() and learning routines.

wlm.cpp 
  - This is a simple test program for testing wlm.hpp routines

project3.cgi: 
  - This will be the cgi which is called from the web. It
    integrates sessions, games, and provides the API to the client.

Pente.hpp: 
  - This describes the game configuration and interfaces with
    the learning machine by providing an interface to state(). We provide
    a make_move() function by passing in the Weights object as a
    reference. (Because Vhat is defined as Weights::Vhat().) 
  - All Xi variables are defined here in Pente::toState()

pente.cpp
  - This is the set of test-routines for Pente.hpp. If we are correct,
    this defends against regressions in Pente.hpp.

shgame.cpp
  - A simple utility to display the state of a game file (passed as a
    cli parameter). Useful for debugging game idiosyncracies.

trainer.cpp
  - The training tool, runs a two opposing computer instances and 
    adjusts weights based on the resulting gametrace. compile with
    '-O3' option for maximum performance.  

batch_train.sh
  - Shell script to run trainer 1500 times.

-=HTML Client=-
  index.html 
  - Simple html.
  
  js/ajax.js
  - All API and AJAX calls happen here.
 
  js/pente.js
  - Routines to draw the board (as well as defining CGI_URL) occur here.

  css/*.css
  - Styles

  images/*.gif
  - Stolen images from pente.net

- games are stored in GAMES_DIRECTORY (defined in wlm.hpp) as "games/"
- "weights.txt" stores the persistent 'weights'. It is defined in
   wlm.hpp