Reproduce “Approximate Optimal Active Learning of Decision Trees” 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

Zunchen Huang, Chenglu Jin

The SAMCAL framework reconstructs unknown binary decision trees from black-box access using a novel approach that combines symbolic encoding with approximate model counting for query selection. This strategy efficiently narrows down hypothesis spaces up to 10^15, converging to the target tree with a minimal number of queries while ensuring robust termination through functional equivalence checks.

Key insights

  • Symbolic representation of decision tree hypothesis spaces combined with approximate model counting enables efficient reasoning over vast combinatorial spaces beyond explicit enumeration.
  • Guiding active query selection by estimating version space reduction with approximate model counting leads to near-optimal logarithmic query complexity.
  • Functional equivalence checks provide a crucial mechanism for reliable termination, especially when approximate model counting's granularity is insufficient to distinguish functionally identical hypotheses.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.