A collection of search space for the DeepHyper package.
Graphviz.
Generate a neural architecture space for fully connected networks with residual connections:
from deepspace.tabular import DenseSkipCoSpace
def create_search_space(input_shape=(54,), output_shape=(7,), **kwargs)
return DenseSkipCoSpace()(input_shape, output_shape, num_layers=10, dropout=0.0)
Generate a neural architecture space for AutoEncoder guided by an estimator:
from deepspace.tabular import SupervisedRegAutoEncoderSpace
factory = SupervisedRegAutoEncoderSpace()(
input_shape=(100,), output_shape=[(100), (10,)]
)