Reproduce “Estimating Implicit Expected Log Density Ratios” 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
Avi Amalanshu, Chandler Squires, Yizhou Lu, Pradeep Ravikumar
The authors introduce a triangular path-based estimator for expected log density ratio (ELDR), a scalar comparing two distributions under a third sampling distribution. It includes samples from that scored distribution in its training path. On semi-synthetic digit data at the hardest mixture setting, median absolute error is 1.36 versus 27.0 for the base method; finite samples from the target still limit estimation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.