Neural Network training, constraint solving and sudoku to connect the two

Notebooks for a hands-on session on training a digit classifier and using it to solve visual sudoku's.

p0: to help you test that all necessary software is installed
p1: first part, training an mnist digit recognizer, predicting and solving a sudoku grid
p2: second part, predicting the grid probabilities and solving a maximum-likelihood visual sudoku

Installation instructions:

if (you do not have 'conda' or 'pip' on your machine):
-> install miniconda as per https://docs.conda.io/en/latest/miniconda.html
if (yo do not have 'classic Jupyter Notebook' on your machine):
-> 'conda install -c conda-forge notebook', see https://jupyter.org/install
download the p0 file to a folder, run 'jupyter notebook' in that folder, open p0, read it and test...

Each of the notebooks can be run as-is, but they don't do meaningful things yet.

The proposed approach is to read and execute it cell-by-cell, modifying parts if you want to experiment or understand better.

At some point, you will encounter a Task: description which challenges you to do or add a missing component.

Have Fun!

Feedback welcome! (just don't push the solutions if you fork the project)

Created by the VUB Data Lab team of Prof. Tias Guns tias.guns@vub.be