algo/improvement: tracking losses for algorithms

[sameerashahh]

Is your request related to a problem? Please describe.
When returning the training and/or validation loss, there is an inconsistency in naming the variables.

Describe the improvement you are suggesting
"train_loss", "val_loss" (if exists), should be the naming convention for all the algorithms when returning the losses dictionary.
aug_ode, dmd, dnn_keras, lr, ode, need to be changed.