Reproduce “Probabilistic computation and uncertainty quantification with emerging covariance” 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
Hengyuan Ma, Yang Qi, Li Zhang, Wenlian Lu, Jianfeng Feng
The moment neural network (MNN) propagates activity means and covariances to produce predictions and an uncertainty signal. Its supervised-mean, unsupervised-covariance (SMUC) rule trains predictions through the mean while covariance changes through nonlinear coupling, without direct covariance supervision. On tested classification tasks, covariance-derived entropy is higher for errors and separates notMNIST inputs, but this does not establish calibrated uncertainty detection.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.