In the .ipynb notebook file in this repository, Google's ML workbench is used to demonstrate how to reproduce some results of Mourrain et. al. from the paper Determining the number of real roots of polynomials through neural networks. That paper demonstrated that neural networks could be used to count the number of real solutions to polynomial systems. For details, see the slides at the link below. They are from an informal talk I gave as a part of the QTM chalk talk series. Keep posted for updates. As a next step, it would be interesting utilize large training sets and then compare with existing solvers.
http://slides.com/jeremyjacobson/deep-learning
If you want to run the code in the notebook here, you must open it using Google's Datalab as it uses import google.datalab.contrib.mlworkbench.commands Here is a direct link for more about Datalab