added in_memory loading to reduce disk reads and increase speed

add testing notebook for convolutional neural network

cleanup to make ready as submodule