## Hacking Tensorflow into a sudoku solver

Google’s Tensorflow has seen great success in deep learning applications. How can we repurpose its functionality for mixed-integer optimization? Specifically, can we use its optimization capabilities to solve a sudoku?