27871.html)

Neural network for handwritten digits recognition; inspiration provided by http://neuralnetworksanddeeplearning.com/index.html

In order to increase isolation and substitulability, the whole project is partitioned in packages.

Particular mentions are needed for the two launch scripts, ‘launch.ipynb’ is extremely helpful for users that are familiar with jupyter notebook; otherwise just make use of ‘launch.py’.

The main workload is performed by the optimisation functions contained in ‘optimisation_functions.py’: they are in charge of starting the learning process taking as input a neural network already initialised (‘neural_network.py’).

For understanding purposes, i suggest starting to read ‘launch.py’ in order to get familiar with the structure model i’ve conceived.