Reproduce “Noise-shaped Sampling and Annealing for Scalable On-the-edge Training of Binary and Ternary AI Models” 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
Zihao Chen, Zhili Xiao, Rui-Jie Zhu, Yipeng Wang, Jason Eshraghian, Shantanu Chakrabartty
The Sigma-Delta Sampler trains discrete binary or ternary weights directly, feeding the difference between gradient signals and realized updates into an accumulator that guides annealed weight flips. On a planted binary optimization problem, it recovered the known global minimum in about 100 steps. A ternary VGG reached 92.20% CIFAR-10 test accuracy, below the 92.88% full-precision baseline; activations remain unquantized.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.