Backpropagation-free Deep Learning for Time-Series Classification
An analysis framework is proposed that relates the per-dataset gap to InceptionTime to dataset properties and tests convolutional feature learning by replacing the flatten readout with global average pooling, measured as a difference-in-differences (DD) against a backpropagation control on the identical architecture.
Miha Kapš, Blaž Bertalanič, Carolina Fortuna
· 0 citations