Reproduce “Continual Learning of Dynamical Systems in Recurrent Neural Networks through Recyclable Unit Gating” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sima Hashemi, Daniel Durstewitz, Georgia Koppe
CRUG lets a recurrent model learn dynamical systems sequentially by gating compact unit sets, freezing selected units, recycling unused units, and allowing one-way connections for forward transfer. On a 160-unit AL-RNN, it achieved median state-space divergence of 1.26 versus 3.81 for CLNP and committed 63 versus 123.5 units. CRUG completed all 10 runs; transfer helped for related, not all, systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.