I need to develop an Adaptive Resonance Theory (ART)-based neural network for research.
Is there interest in adding ART-based neural networks to PyTorch?
This is a family of networks that implement the adaptive resonance theories of cognition (Carpenter and Grossberg). They avoid the “catastrophic loss” issue that exist with other network architectures, i.e., they do not have to be completely re-trained from scratch if a new pattern is uncovered but can add this pattern on-the-fly.