Creates an SGD optimizer with optional momentum.
Learning rate
Momentum factor (default: 0.0)
Weight decay coefficient (default: 0.0)