Author: Jo Schlemper, Imperial College London 
4th Year Project


====================================================================
1. In order to run the program, you will need the following packages:

theano, opencv, sklearn, numpy, scipy, matplotlib, cPickle and Image


====================================================================
2. The directory contains 3 experiments described in the report:

noise_classification.py - Classify happy/sad face using models trained
on different ratio of training data

associate_digits.py - Learn even-odds of handwritten digits

associate_kanade.py - Attachment Theory simulation


====================================================================
3. Brief Overview of the directories:

models - contains all the models, including RBM, DBN and AssociativeDBN

examples - contains example files of how to initialise, run, store the models 

test - all the test cases

kanade_preprocessing - methods used to crop/preprocess face images

result - all the logs/results from running the experiments. Also,
methods to plot these files are included. All the images will be saved
in ``data'' directory.


====================================================================
4. Dataset

You will need:
MNIST handwritten digits - mnist.pkl.gz 
Cohn Kanade face images - stored in /data/ using cPickle 

These data are loaded using mnist_loader.py and kanade_loader.py
respectively.