EPFL Neuro Symposium // Surprise, Curiosity and Reward: from Neuroscience to AI

Thumbnail

Event details

Date 08.02.2021 09.02.2021
Hour 14:0019:00
Speaker See list below
Location
Online
Category Conferences - Seminars
This event has been recorded and is available on-line : Reward is key in models of reinforcement learning. However, humans also learn in the absence of reward, for example, driven by curiosity, e.g., we acquire knowledge about the lay-out of a city by walking around. But then the question arises: can we integrate curiosity-driven behavior into the framework reinforcement learning? Is curiosity the same as novelty-seeking? Or is it surprise-seeking? Is intrinsic motivation just an internal reward or something else?

In neuroscience and psychology, reinforcement learning algorithms are used to explain human brain activity and behavior while in artificial intelligence, they are used to learn to play complex games (amongst other applications). This workshop brings together researchers from both neuroscience and AI, since notions of curiosity and surprise have been used in both domains to address learning in volatile environments and in tasks where reward is sparse and of unknown value.

List of speakers (in alphabetical order): Commitee members
  • Wulfram Gerstner, Computational Neuroscience, EPFL
  • Michael Herzog, Psychophysics Laboratory, EPFL
  • Johanni Brea, Computational Neuroscience, EPFL
  • Mackenzie Mathis, Adaptative Motor Control, EPFL
  • Alexander Mathis, Computational neuroscience & AI,  EPFL
Moderators
  • Michael Herzog, Psychophysics Laboratory, EPFL
  • Johanni Brea, Computational Neuroscience, EPFL
  • Mackenzie Mathis, Adaptative Motor Control, EPFL
Q&A with information/links from Monday 8 February talks: BMI_08.02_Q&R

Links

Practical information

  • Informed public
  • Registration required

Organizer

  • Michael Herzog, Psychophysics, School of Life Science, Brain Mind Institute, EPFL & Wulfram Gerstner , Computational Neuroscience, School of Computer Science and Life Sciences, EPFL

Share