Clip gradient to the range of [-clip_gradient, clip_gradient] If clip_gradient <= 0, gradient clipping is turned off. grad = max(min(grad, clip_gradient), -clip_gradient).
If true, lazy updates are applied if gradient's stype is row_sparse and both weight and momentum have the same stype
The decay rate of momentum estimates at each epoch.
Rescale gradient to grad = rescale_grad*grad.
Weight decay augments the objective function with a regularization term that penalizes large weights. The penalty scales with the square of the magnitude of each weight.