Balduzzi, D; Tuyls, K; Pérolat, J; Graepel, T; (2018) Re-evaluating evaluation. In: Bengio, S and Wallach, HM and Larochelle, H and Grauman, K and Cesa-Bianchi, N and Garnett, R, (eds.) Proceedings of the 32nd Conference on Neural Information Processing Systems (NeurIPS 2018). (pp. pp. 3272-3283). Neural Information Processing Systems Foundation, Inc.: Montréal, Canada.

Preview

Text 7588-re-evaluating-evaluation.pdf - Published Version Download (540kB) | Preview

Abstract

Progress in machine learning is measured by careful evaluation on problems of outstanding common interest. However, the proliferation of benchmark suites and environments, adversarial attacks, and other complications has diluted the basic evaluation model by overwhelming researchers with choices. Deliberate or accidental cherry picking is increasingly likely, and designing well-balanced evaluation suites requires increasing effort. In this paper we take a step back and propose Nash averaging. The approach builds on a detailed analysis of the algebraic structure of evaluation in two basic scenarios: agent-vs-agent and agentvs-task. The key strength of Nash averaging is that it automatically adapts to redundancies in evaluation data, so that results are not biased by the incorporation of easy tasks or weak agents. Nash averaging thus encourages maximally inclusive evaluation – since there is no harm (computational cost aside) from including all available tasks and agents.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item