A comparison of linear and deep learning based models for dimensionality reduction on MNIST data
MNIST training dataset
The code summarises the dataset into 64 dimensions using autoencoders and principle component analysis and then compares the results based on the reconstruction (decoders) of the images. It is observed that deep learning based autoencoders summarise the information better as compared to the linear PCA model.
execute the Auto_Encoders-vs-PCA.py