Prof. Ronald de Wolf
CWI, University of Amsterdam, and Google Quantum AI
Venue: Wave Auditorium, Masdar Office
24th November 2025, 10:00AM - 11:00AM (GST)
| Title: | Quantum algorithms for boosting in machine learning |
| Abstract: | Boosting is a beautiful technique in machine learning that allows to run a "weak learner" (a learning algorithm which produces hypotheses that are only slightly better than random) multiple times to obtain a "strong learner" (which produces hypotheses with small generalization error). The most famous classical boosting algorithm is Freund and Schapire's Adaboost from the 1990s. Arunachalam and Maity (ICML'20) gave a quantum version of Adaboost which is faster in some parameters but worse in others. This talk presents two improved quantum boosting algorithms: one due to Izdebski and myself (ESA'23) which quantizes Servedio's classical Smoothboost algorithm, and one due to Abbas, Chen, Nguyen, and myself (arXiv:2510.05089) which quantizes Kale's version of a smooth booster, based on approximate Bregman projections. Our second algorithm is the first quantum boosting algorithm that is better than the best classical one in some parameters, and no worse in all others. |
| Bio: | Ronald de Wolf (1973) studied computer science and philosophy at the Erasmus University Rotterdam, with a focus on logic-based machine learning. He obtained his PhD in 2001 from the University of Amsterdam and CWI on a thesis about quantum computation and communication complexity, advised by Harry Buhrman and Paul Vitanyi. Subsequently he spent a postdoctoral year at UC Berkeley. Currently he is a senior researcher at CWI and full professor at the University of Amsterdam. |