Full steam ahead! Unraveling Forward-Forward Neural Networks (Ep. 232)
Data Science at Home
00:00
The Backward Propagation-Based Neural Network
Back propagation is a trick that we use from a computational perspective in order to speed up the learning phase. In biological brains, it would be very inefficient to implement some kind of back-propagation system. Back propagate requires usually perfect knowledge of the computation performed in the forward pass and can only work with black boxes or models which are not differentiable. Theforward forward algorithm plays in line with what a normal brain would work; there's no back propagator at all.
Transcript
Play full episode
