High quality decisions from conflicting advice
The Netherlands Organisation for Scientific Research (NWO) has awarded a Rubicon grant to Wouter Koolen from the Centrum Wiskunde & Informatica (CWI). He receives the grant for his research project ‘Game-Theoretically Optimal Online Learning: From Conflicting Advice to High-Quality Decisions'. The Rubicon program aims to encourage talented researchers and offers them to gain experience at a top research institute outside the Netherlands.
Online learning
The focus of Koolen’s reserach is online learning, an emerging discipline on the interface between computer science, information theory and statistics. Online learning studies recurring decision problems with conflicting expert advice. Every day we base many practical decisions and actions on opinions of experts. Unfortunately, the experts often disagree.
The best expert
Where do you find the most reliable weather forecast? Which planner chooses the fastest way home? Who provides the best advice on stock prices? For a decision on a financial stock exchange transaction you would rely on expert advice. However, it is hard to make a high quality decision when various experts provide different information. An additional problem is that the relative performance of experts may also change over time.
New computer programs
The new computer programs that Koolen will develop automatically learn to make high-quality decisions. The programs execute the optimal strategy, in which the opinions of all available experts are combined. The computer programs of Koolen also know exactly when to switch between different experts over time.
Royal Holloway
Koolen is going to do his research at the Computer Learning Research Centre at Royal Holloway, University of London. The centre is internationally recognized for its research into machine learning problems. The research group is led by Professor Vovk, an internationally recognized leading authority in the field of machine learning and online learning in particular.
