A Parallel version of Stochastic Gradient Descent as detailed in: Parallelized Stochastic Gradient Descent .
Takes advantage of all available cores locally and on remote servers.
Example usage (linear regression):
everywhere using PSGD, Base.Test
x = linspace(1, 100, 100000)
y = linspace(100, 200, 100000)
res = psgd(.0002, 200, x, y, lineargradient)
