Deep neural network fit to a dataset of textual descriptors of different neural network architectures (for CS 446)
Each architecture was trained on the Cifar-10 dataset and represent error for training different models for image detection.
Features/labels:
>200 features,
and two sets of labels.
Deep neural network architecture:
2 hidden layers,
a sigmoid activation function,
batch normalization,
Gradient method: Adam, learning rate = 0.01
Loss Function: SmoothL1Loss