Unsupervised Feature Learning and Deep Learning Tutorial

Implement the exercises of UFLDL Tutorial with python 3

Tutorial Website: http://ufldl.stanford.edu/wiki/index.php/UFLDL_Tutorial

Packages required

• Numpy
• SciPy
• matplotlib
• scikit-learn

Sparse Autoencoder

• load_MNIST.py: Load MNIST dataset
• sample_images.py: Load sample images for testing sparse autoencoder
• sparse_autoencoder.py: Functions used in sparse autoencoder
• train.py: Train sparse autoencoder on sample images
• train_mnist.py: Train sparse autoencoder with MNIST data
• check_numerical_gradient.py: Check numerical gradients
• display_network.py: Display visualized features

Preprocessing: PCA & Whitening

• pca_2d.py: PCA, PCA whitening and ZCA whitening in 2D
• pca_gen.py: PCA and Whitening on natural images

Softmax Regression

• softmax.py: Functions used in softmax regression
• softmax_exercise.py: Classify MNIST digits

Self-Taught Learning and Unsupervised Feature Learning

• stl_exercise.py: Classify MNIST digits with self-taught learning and softmax regression

Building Deep Networks for Classification

• stacked_autoencoder.py: Functions used in stacked autoencoder
• stacked_autoencoder_exercise.py: Use a stacked autoencoder for digit classification

Linear Decoders with Autoencoders

• linear_decoder_exercise.py: Implement a linear decoder and apply it to learn features on color images

Working with Large Images (Convolutional Neural Networks)

• cnn.py: Functions used in convolution neural networks
• cnn_exercise.py: Classify STL-10 images