Matlab implementation of convolutional auto-encoders, based on paper Stacked Convolutional Auto-Encoders for Hierarchical Feature Extraction compatible with DeepLearnToolbox https://github.com/rasmusbergpalm/DeepLearnToolbox
test.m is an example of how to set up and train a convolutional auto-encoder, visualize the first layer kernels and reconstruction results alongside the original input, use the training result to initialize a convolutional neural network with the same architecture, and compare error rate with random initialization.
cae_check_grad method in cae_train.m can be turned on to verify the gradients numerically.
An example of 24 first layer convolution weights trained on the KITTI image set
