ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 1906.11385
34
6
v1v2 (latest)

A Tight Analysis of Greedy Yields Subexponential Time Approximation for Uniform Decision Tree

26 June 2019
Ray Li
Percy Liang
Stephen Mussmann
ArXiv (abs)PDFHTML
Abstract

Decision Tree is a classic formulation of active learning: given nnn hypotheses with nonnegative weights summing to 1 and a set of tests that each partition the hypotheses, output a decision tree using the provided tests that uniquely identifies each hypothesis and has minimum (weighted) average depth. Previous works showed that the greedy algorithm achieves a O(log⁡n)O(\log n)O(logn) approximation ratio for this problem and it is NP-hard beat a O(log⁡n)O(\log n)O(logn) approximation, settling the complexity of the problem. However, for Uniform Decision Tree, i.e. Decision Tree with uniform weights, the story is more subtle. The greedy algorithm's O(log⁡n)O(\log n)O(logn) approximation ratio was the best known, but the largest approximation ratio known to be NP-hard is 4−ε4-\varepsilon4−ε. We prove that the greedy algorithm gives a O(log⁡nlog⁡COPT)O(\frac{\log n}{\log C_{OPT}})O(logCOPT​logn​) approximation for Uniform Decision Tree, where COPTC_{OPT}COPT​ is the cost of the optimal tree and show this is best possible for the greedy algorithm. As a corollary, we resolve a conjecture of Kosaraju, Przytycka, and Borgstrom. Leveraging this result, for all α∈(0,1)\alpha\in(0,1)α∈(0,1), we exhibit a 9.01α\frac{9.01}{\alpha}α9.01​ approximation algorithm to Uniform Decision Tree running in subexponential time 2O~(nα)2^{\tilde O(n^\alpha)}2O~(nα). As a corollary, achieving any super-constant approximation ratio on Uniform Decision Tree is not NP-hard, assuming the Exponential Time Hypothesis. This work therefore adds approximating Uniform Decision Tree to a small list of natural problems that have subexponential time algorithms but no known polynomial time algorithms. All our results hold for Decision Tree with weights not too far from uniform. A key technical contribution of our work is showing a connection between greedy algorithms for Uniform Decision Tree and for Min Sum Set Cover.

View on arXiv
Comments on this paper