Table 2 FCNN training parameters used for each classification problem.
From: Decoding behavior from global cerebrovascular activity using neural networks
Sleep/wake | Movement | ||
|---|---|---|---|
Hidden layers | 1 | ||
Hidden layer neurons | 3 | ||
Output neurons | 4 | 2 | |
Output activation | Softmax | ||
Max epochs | 10,000 (+ early stopping, cf. text) | ||
Optimization | Minibatch stochastic gradient descent | ||
Minibatch size | 8 | 6 | |
Initial learning rate | 0.05 | 5 | |
Learning rate decay | \(\alpha =\frac{\alpha }{1+ep}\) | ||
Regularization | None | ||
