Parallelizing Stochastic Gradient Descent for Least Squares Regression: mini-batching, averaging, and model misspecification | Read Paper on Bytez