Adaptive Learning Rate Methods

YOGI

Variant of Adam that uses an adaptive variance controller to stabilize training, particularly effective when data has non-stationary distributions or noisy gradients.

← Wstecz