Karl Friston: Active inference and artificial curiosity
Professor Karl Friston (University College London) speaks at the University of Edinburgh, 31st March 2017.
Abstract: This talk offers a formal account of insight and learning in terms of active (Bayesian) inference. It deals with the dual problem of inferring states of the world and learning its statistical structure. In contrast to current trends in machine learning (e.g., deep learning), we focus on how agents learn from a small number of ambiguous outcomes to form insight.
Karl will present simulations of abstract rule-learning and approximate Bayesian inference to show that minimising (expected) free energy leads to active sampling of novel contingencies. This epistemic, curiosity-directed behaviour closes ‘explanatory gaps’ in knowledge about the causal structure of the world; thereby reducing ignorance, in addition to resolving uncertainty about states of the known world.
We then move from inference to model selection or structure learning to show how abductive processes emerge when agents test plausible hypotheses about symmetries in their generative models of the world. The ensuing Bayesian model reduction evokes mechanisms associated with sleep and has all the hallmarks of ‘aha moments’.
Видео Karl Friston: Active inference and artificial curiosity канала Psychology at the University of Edinburgh
Abstract: This talk offers a formal account of insight and learning in terms of active (Bayesian) inference. It deals with the dual problem of inferring states of the world and learning its statistical structure. In contrast to current trends in machine learning (e.g., deep learning), we focus on how agents learn from a small number of ambiguous outcomes to form insight.
Karl will present simulations of abstract rule-learning and approximate Bayesian inference to show that minimising (expected) free energy leads to active sampling of novel contingencies. This epistemic, curiosity-directed behaviour closes ‘explanatory gaps’ in knowledge about the causal structure of the world; thereby reducing ignorance, in addition to resolving uncertainty about states of the known world.
We then move from inference to model selection or structure learning to show how abductive processes emerge when agents test plausible hypotheses about symmetries in their generative models of the world. The ensuing Bayesian model reduction evokes mechanisms associated with sleep and has all the hallmarks of ‘aha moments’.
Видео Karl Friston: Active inference and artificial curiosity канала Psychology at the University of Edinburgh
Показать
Комментарии отсутствуют
Информация о видео
5 апреля 2017 г. 13:31:48
00:57:41
Другие видео канала
Free Energy Principle — Karl FristonDynamic Causal Modelling - Karl FristonProvost’s Lecture: Karl Friston on “I Am Therefore I Think”Karl Friston - Prediction: How Forecasting and Prospection Shape ThoughtI am therefore I think by Karl FristonBrain Imaging Karl FristonLife beyond Psychology at Edinburgh- Graduate Stories: Jining ZhangA visual guide to Bayesian thinkingUnderstanding Artificial Intelligence and Its Future | Neil Nie | TEDxDeerfield#033 Karl Friston - The Free Energy PrincipleEmbodied Cognition Karl FristonDysconnection Hypothesis of Schizophrenia Karl FristonThe Deep End of Deep Learning | Hugo Larochelle | TEDxBostonBI 028 Sam Gershman: Free Energy Principle & Human MachinesConsciousness and the Bayesian brain - 08 - Sandler Conference 2014Karl Friston on What Is Life, Consciousness, the meta-Hard Problem, and the Free Energy PrincipleMindscape 87 | Karl Friston on Brains, Predictions, and Free EnergyKarl Friston - 2016 CCN Workshop: Predictive CodingExploring the Crossroads of Attention and Memory in the Aging Brain: Views from the InsideHow Generative Models Work