AdaBoost Does Not Always Cycle: A Computer-Assisted Counterexample

2026-04-08

AI Summary

This study resolves a long-standing open question regarding whether AdaBoost necessarily converges to a finite cycle under exhaustive training. By constructing a counterexample based on a block-product structure—combining two branch maps, each exhibiting a period-2 orbit but whose linearized return maps have dominant eigenvalues with an irrational logarithmic ratio—the authors demonstrate that the sequence of weak learners selected by AdaBoost can display asymptotic frequencies that are irrational, thereby precluding eventual periodicity. This work presents the first rigorously verified instance of non-periodic AdaBoost dynamics, refuting the conjecture that AdaBoost must always converge to a cyclic behavior, and introduces an irrational-frequency mechanism to explain such phenomena. All results are formally validated through a combination of exact rational arithmetic, dynamical systems analysis, and symbolic computation.

Technology Category

Machine Learning: Online Learning & BanditsGame Theory and Economic Paradigms: Adversarial LearningCognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingGraph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphsUser Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

We give a computer-assisted counterexample to the open question, posed by Rudin, Schapire, and Daubechies in COLT 2012, of whether exhaustive AdaBoost always converges to a finite cycle. The construction is based on a block-product gadget whose two factors share an exact period-2 orbit for their 5-step branch maps, but whose linearized return maps have dominant eigenvalues with an irrational logarithmic ratio. This irrationality forces the burst-winner sequence to have an irrational asymptotic frequency, precluding eventual periodicity. All assertions are certified by exact rational arithmetic. This work was developed in collaboration with GPT-5.4 Pro and Claude Opus 4.6.

Problem

• AdaBoost
• convergence
• periodicity
• counterexample
• machine learning

Innovation

• AdaBoost
• non-periodicity
• computer-assisted proof
• block-product gadget
• irrational eigenvalue ratio