Use of two optimizers?

[lucvanwyk]

Is their a reason you're using two/three separate optimizers in learning? Will it make a difference if the parameters of all models (cnn/mlps) were given to a single optimizer?