Reproduce “Accelerated physical emulation of Bayesian inference in spiking neural networks” 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
Akos F. Kungl, Sebastian Schmitt, Johann Klähn, Paul Müller, Andreas Baumbach, Dominik Dold, Alexander Kugele, Nico Gürtler and 16 more
The authors demonstrate sampling-based inference on BrainScaleS: spiking neurons represent guesses, learned weights compensate for analog variability, and an on-chip recurrent network supplies noise. On reduced, binarized four-class MNIST, classification error was 4.45% versus 3.89% for an RBM; occluded-pixel error reached about 10% within 50 milliseconds. The measured speed-up was 210×, with host-based learning and no energy or full-MNIST evaluation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.