Beyond Shortcuts: Unlocking true Attractor Dynamics in Neural Networks

New research reveals the challenges continuous attractor networks face in learning meaningful transitions, and how geometric constraints can limit their ability to represent complex environments.







