Deep, Resonant Neural Nets

In the 1980’s Carpenter and Grossberg of Boston University advanced the idea of “Adaptive Resonance Theory” (ART). Carpenter and Grossberg presented significant papers at various Neural Net conferences, but now ART is mostly forgotten. ART proposes a very simple learning algorithm that has not been applied to deep nets, until now. Practically all deep neural nets currently implemented use another learning algorithm, called backpropagation (backprop).

All implementations of ART (and ART2, ART3 later) in the 1980s are at most three levels deep. Centerus is building a deep, ART-like net.