optimizer

• method: Adam (default) / Adadelta, Adagrad, AdamW, Adamax, ASGD, SGD, RMSprop, Rprop

---

optimizer determines the optimization method. The usage of optimizer is as below. SIMPLE-NN supports Adam, Adadelta, Adagrad, AdamW, Adamax, ASGD, SGD, RMSprop and Rprop.

# input.yaml
neural_network:
    optimizer:
        method: Adam
        params:

As shown below, in general, Adam type optimizer shows the best convergence.