Reproduce “Understanding Sample Efficiency in Predictive Coding” 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
Gaspard Oliviers, Elene Lominadze, Rafal Bogacz
The authors analyze predictive coding (PC) and backpropagation (BP) updates in deep linear networks. They show weight-dependent transformations can redirect BP updates, while PC can counter them; specified activity-based rescaling gives perfect alignment for single samples and under a batch-size condition. Experiments find PC reaches lower error in fewer training steps, not necessarily less computation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.