Reproduce “Hypnos: A Dual-Phase End-to-End Platform for Scalable Continual Learning at the Edge” 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
Guilherme Guedes, Marco Apolinario Lainez, Charlotte Frenkel
Hypnos divides continual learning between an edge device and a charger: the device makes gradient-free prototype updates and stores compressed latent codes, while the charger uses replay and gradient-based consolidation. On CORe50, its INT8 A8 system reaches 78.6% final accuracy with wake-phase learning fully on-chip. This result covers image classification on the benchmark, not a deployed wearable.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.