BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//TECoSA - ECPv6.17.1//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:TECoSA
X-ORIGINAL-URL:https://www.tecosa.center.kth.se
X-WR-CALDESC:Events for TECoSA
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:UTC
BEGIN:STANDARD
TZOFFSETFROM:+0000
TZOFFSETTO:+0000
TZNAME:UTC
DTSTART:20240101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=UTC:20250206T150000
DTEND;TZID=UTC:20250206T160000
DTSTAMP:20250203T100948Z
CREATED:20250117T091105Z
LAST-MODIFIED:20250203T100948Z
UID:7586-1738854000-1738857600@www.tecosa.center.kth.se
SUMMARY:TECoSA Seminar - Real-Time Learning\, Planning and Control with Curiosity-Driven Exploration
DESCRIPTION:We aim to bring you a TECoSA Seminar on the first Thursday of each month during term-time.  All are welcome to attend and we look forward to some lively discussions. Members can accept the Outlook invitations\, non-members can email tecosa-admin@kth.se to register.\nOur February seminar is with Patrick Hammer\, AI researcher and postdoctoral scholar in the Division of Robotics\, KTH. The seminar will take place at Gladan\, KTH. Register for the seminar here. \nIt is also possible to attend the seminar via Zoom. \nAbstract: This talk delves into enabling embodied agents to handle discrete and continuous domains with high-dimensional input using a unified model\, eliminating the need for hyperparameter tuning. We explore recent advancements in Model-Based Deep Reinforcement Learning and revisit the foundational goals of AI and Cybernetics: creating general-purpose adaptive systems. The discussion emphasizes learning efficiency through inductive biases\, causal reasoning and curiosity-driven exploration\, showcasing an experiential learning technique with unparalleled sample efficiency in grid world environments. Comparative results on the Minigrid benchmark and preliminary experiments with a simulated mobile robot highlight the approach’s versatility. The talk concludes with insights into unifying diverse methodologies and an optimistic outlook for advancing research toward AI’s original\, ambitious objectives. \nBio: Patrick Hammer is an AI researcher and postdoctoral scholar in the Division of Robotics\, Perception\, and Learning at KTH. His research focuses on advancing the autonomy of intelligent agents\, especially mobile robots\, by providing them with higher levels of adaptiveness\, emphasizing the integration of Non-Axiomatic Reasoning and Deep Neural Networks. His long-term vision is to replicate\, in computer systems\, key aspects of natural intelligence as showcased by higher-developed animals in their cognitive capacities and ability to improvise.
URL:https://www.tecosa.center.kth.se/event/tecosa-seminar-real-time-learning-planning-and-control-with-curiosity-driven-exploration/
CATEGORIES:Seminar
END:VEVENT
END:VCALENDAR