Hi, it seems the python code wrapper has support for only the Feed forward NNs correct? such as this
from stg import STG
model = STG(task_type='regression',input_dim=X_train.shape[1], output_dim=1, hidden_dims=[500, 50, 10], activation='tanh', optimizer='SGD', learning_rate=0.1, batch_size=X_train.shape[0], feature_selection=True, sigma=0.5, lam=0.1, random_state=1, device="CPU")
or does it support custom NN architectures?? such as a CNN layer somewhere in between the hidden layers for example? Thanks :)
Hi, it seems the python code wrapper has support for only the Feed forward NNs correct? such as this
from stg import STG
model = STG(task_type='regression',input_dim=X_train.shape[1], output_dim=1, hidden_dims=[500, 50, 10], activation='tanh', optimizer='SGD', learning_rate=0.1, batch_size=X_train.shape[0], feature_selection=True, sigma=0.5, lam=0.1, random_state=1, device="CPU")
or does it support custom NN architectures?? such as a CNN layer somewhere in between the hidden layers for example? Thanks :)