Reproduce “Conditional Distribution Quantization in Machine 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
Blaise Delattre, Sylvain Delattre, Alexandre Vérine, Alexandre Allauzen
Conditional Competitive Learning Vector Quantization (CCLVQ) learns several input-dependent candidate outputs and estimates a weight for each. It assigns each training target to the lowest-loss candidate, updates that expert, and trains a classifier on assignments. On CIFAR-10 BigGAN, two experts report FID 7.34 ± 0.03 and Recall 0.72 ± 0.01; these results do not establish calibrated uncertainty or general gains.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.