- This event has passed.
Manuel Lopes (hosted by Marc Toussaint): Optimal Behavior Without Optimal Rewards : Artificial Vs Natural
18 June, 2020 @ 10:00 am - 11:00 am
Abstract:
Research in robotics and A.I. aims at optimizing very specific task rewards. Intelligent animals have a high degree of curiosity, and recent
Research in robotics and A.I. aims at optimizing very specific task rewards. Intelligent animals have a high degree of curiosity, and recent
results have shown that instrumental reward optimization is a poor explanation for their behavior. We can show that to explain empirical
results from animals, we need to have the drive to optimize reward, a drive to reduce uncertainty, and a drive for positive cues. We then show examples in robotics where a more complex reward system provides benefits in learning.
References:
Daddaoua, N., Lopes, . & Gottlieb, J. Intrinsically motivated oculomotor exploration guided by uncertainty reduction and conditioned
reinforcement in non-human primates. Sci Rep 6, 20202 (2016). https://doi.org/10.1038/srep20202
Lopes, M., Lang, T., Toussaint, M., & Oudeyer, P. Y. (2012). Exploration in model-based reinforcement learning by empirically estimating learning progress. In Advances in neural information processing systems (pp. 206-214).
***Want to know more about this lecture? Contact us at communication@scioi.de
(Photo by Franck V. on Unsplash)
Marchstr. 23 – MAR 5-2
10587 Berlin
Email: info@scioi.de
Tel: +49 30 314 77344
