jeromeku / convolutional-neural-network Goto Github PK

-> This is a solution to the Convolutional Neural Network exercise in the Stanford UFLDL Tutorial(http://ufldl.stanford.edu/wiki/index.php/Exercise:Convolution_and_Pooling)
-> The code has been written in Python using Scipy, Numpy and Matplotlib
-> The code is bound by The MIT License (MIT)

Running the code:

-> Download the data files: 'stlTrainSubset.mat', 'stlTestSubset.mat', 'opt_param.npy', 'zca_white.npy' and 'mean_patch.npy' and the code file 'convolutionalNeuralNetwork.py'
-> Put them in the same folder, and run the program by typing in 'python convolutionalNeuralNetwork.py' in the command line
-> You should first get an image of the learned Sparse Autoencoder Linear weights as in 'output.png', the code for which is available at the following link : https://github.com/siddharth950/Sparse-Autoencoder-Linear
-> The data files 'opt_param.npy', 'zca_white.npy' and 'mean_patch.npy' were also obtained using the same code
-> This code takes about 8 hours to execute on an i3 processor

Code written by: Siddharth Agrawal
Email ID: [email protected]