Reproduce “Unsupervised Representation Learning for Binary Networks by Joint Classifier Learning” 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
Dahyun Kim, Jonghyun Choi
BURN trains binary networks for self-supervised representation learning, using a pretrained floating-point feature extractor and a classifier trained jointly with the binary network to provide changing prediction targets. A feature-similarity loss helps stabilize training. Across tested downstream tasks, BURN outperforms compared self-supervised binary baselines and sometimes supervised pretraining, but still requires the pretrained floating-point feature extractor.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.