Reproduce “Cycle Consistent Probability Divergences Across Different Spaces” 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
Zhengxin Zhang, Youssef Mroueh, Ziv Goldfeld, Bharath K. Sriperumbudur
The authors introduce UBGM and kernelized GMMD to learn maps in both directions between distributions in different spaces. The maps preserve within-space geometry, match each distribution to the other, and penalize compositions that fail to approximate identity. On synthetic shapes, mappings appear qualitatively aligned. In one biplane experiment, applying trained maps to 8,000 new points per distribution took 63 ms, versus 26 minutes to recompute a GW coupling; this is task-specific inference timing.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.