This report is a description of our approach for the second mini-project in the Deep Learning course given at EPFL, Lausanne. The project aims to create a mini deep learning framework and use it for a simple classification task with randomly generated data.
The provided script is run without argument. Plots and execution time can be enabled by changing two booleans at the beginning of the script. One will enable plots similar to figures 2, 3, 4. The other will compute and display the training time of the network.