Loss Minimization

gradient descent towards consensus