This is a python implementation of parallel computation of stochastic gradient descent. 
This is approach is called Hogwild! based on the following research paper "A Lock-Free Approach to Parallelizing Stochastic Gradient Descent".