A pure implementation for sparse denoising autoencoder with adaptive evolutionary training using Scipy. The sparse implementation makes the algorithm scalable to high dimensional data and trainable on CPUs.