Stanford Unsupervised Feature Learning and Deep Learning Tutorial
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is even with jatinshah:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
output
LICENSE
README.md
cnn.py
cnn_exercise.py
display_network.py
gradient.py
linear_decoder_exercise.py
load_MNIST.py
load_images.py
pca_gen.py
sample_images.py
softmax.py
softmax_exercise.py
sparse_autoencoder.py
stacked_ae_exercise.py
stacked_autoencoder.py
stl_exercise.py
train.py

README.md

Stanford Unsupervised Feature Learning and Deep Learning Tutorial

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

Sparse Autoencoder

Sparse Autoencoder vectorized implementation, learning/visualizing features on MNIST data

Preprocessing: PCA & Whitening

Implement PCA, PCA whitening & ZCA whitening

Softmax Regression

Classify MNIST digits via softmax regression (multivariate logistic regression)

Self-Taught Learning and Unsupervised Feature Learning

Classify MNIST digits via self-taught learning paradigm, i.e. learn features via sparse autoencoder using digits 5-9 as unlabelled examples and train softmax regression on digits 0-4 as labelled examples

Building Deep Networks for Classification (Stacked Sparse Autoencoder)

Stacked sparse autoencoder for MNIST digit classification

Linear Decoders with Auto encoders

Learn features on 8x8 patches of 96x96 STL-10 color images via linear decoder (sparse autoencoder with linear activation function in output layer)

Working with Large Images (Convolutional Neural Networks)

Classify 64x64 STL-10 images using features learnt via linear decoder (previous section) and convolutional neural networks

  • cnn.py: Convolution neural networks. Convolve & Pooling functions
  • cnn_exercise.py: Classify STL-10 images