Preproccessing All image preprocessing can be done in image_manipulation.py and the functions are documented with intended usage. Logistic Regression The logistic regressor is implemented in logistic/logistic_regression.py as the class MulticlassLogisticRegressor. It provides fit and predict method similar to scikit learn classifier, with the exception that predict only works on one example at a time. Neural Network To use a neural network, create the network as follows: import nn net = nn.NeuralNework(num_input_nodes, num_hidden_layers, [list_of_nodes_per_layer], num_output_nodes, learning_rate) nn.fit(xs_np_array, ys_np_array) pred_y = nn.predict(xs_np_array) If you want to run cross validation, simply run nn.cross_validation(). SVM We essentially use scikit-learn's implementation of svm, and all testing was done through interactive python sessions. The final test code for svm can be found in svm.py, if you want to get an idea about how we used svm. CNN Details on running the CNN can be found in the Theano tutorials: http://deeplearning.net/tutorial/lenet.html.