BEGIN:VCALENDAR
PRODID:-//AddEvent Inc//AddEvent.com v1.7//EN
VERSION:2.0
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
BEGIN:STANDARD
DTSTART:20261101T010000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260308T030000
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
DESCRIPTION:Achieving fast and stable off-policy learning in deep reinforcement learning (RL) is chal-\nlenging. Most existing methods rely on semi-gradient temporal-difference (TD) methods for\ntheir simplicity and efficiency\, but are consequently susceptible to divergence. While more\nprincipled approaches like Gradient TD (GTD) methods have strong convergence guarantees\,\nthey have rarely been used in deep RL. Recent work introduced the Generalized Projected\nBellman Error (GPBE)\, enabling GTD methods to work efficiently with nonlinear function ap-\nproximation. However\, this work is only limited to one-step methods\, which are slow at credit\nassignment and require a large number of samples. In this paper\, we extend the GPBE objective\nto support multistep credit assignment based on the λ-return and derive three gradient-based\nmethods that optimize this new objective. We provide both a forward-view formulation com-\npatible with experience replay and a backward-view formulation compatible with streaming\nalgorithms. Finally\, we evaluate the proposed algorithms and show that they outperform both\nPPO and StreamQ in MuJoCo and MinAtar environments\, respectively.\n\nEsraa is a third-year PhD student at the RLAI lab advised by Martha White. Her main research interests lie primarily in online reinforcement learning\, real-time recurrent learning\, and continual learning. She received her M.Sc. in Computing Science from the University of Alberta\, where she worked on efficient real-time recurrent learning approaches for online reinforcement learning. \n\nTo unsubscribe to this calendar please email: cohere-labs-reinforcement-learning+unsubscribe@cohere.com mailto:cohere-labs-reinforcement-learning+unsubscribe@cohere.com\n\n------\n\nCreate your own Add to Calendar links with addevent.com/r/a \n
X-ALT-DESC;FMTTYPE=text/html:Achieving fast and stable off-policy learning in deep reinforcement learning (RL) is chal-<br />lenging. Most existing methods rely on semi-gradient temporal-difference (TD) methods for<br />their simplicity and efficiency, but are consequently susceptible to divergence. While more<br />principled approaches like Gradient TD (GTD) methods have strong convergence guarantees,<br />they have rarely been used in deep RL. Recent work introduced the Generalized Projected<br />Bellman Error (GPBE), enabling GTD methods to work efficiently with nonlinear function ap-<br />proximation. However, this work is only limited to one-step methods, which are slow at credit<br />assignment and require a large number of samples. In this paper, we extend the GPBE objective<br />to support multistep credit assignment based on the λ-return and derive three gradient-based<br />methods that optimize this new objective. We provide both a forward-view formulation com-<br />patible with experience replay and a backward-view formulation compatible with streaming<br />algorithms. Finally, we evaluate the proposed algorithms and show that they outperform both<br />PPO and StreamQ in MuJoCo and MinAtar environments, respectively.<br><br>Esraa is a third-year PhD student at the RLAI lab advised by Martha White. Her main research interests lie primarily in online reinforcement learning, real-time recurrent learning, and continual learning. She received her M.Sc. in Computing Science from the University of Alberta, where she worked on efficient real-time recurrent learning approaches for online reinforcement learning. <br><br>To unsubscribe to this calendar please email: <a href="mailto:cohere-labs-reinforcement-learning+unsubscribe@cohere.com" target="_blank">cohere-labs-reinforcement-learning+unsubscribe@cohere.com</a><br /><br />------<br /><br />Create your own Add to Calendar links with addevent.com/r/a <br>
UID:e3c8f54d93ed4cf7be5a77a938b58f4daddeventcom
SUMMARY:Esraa Elelimy - Deep Reinforcement Learning with Gradient Eligibility Traces (RL)
DTSTART;TZID=America/Los_Angeles:20250929T090000
DTEND;TZID=America/Los_Angeles:20250929T100000
DTSTAMP:20260826T072157Z
TRANSP:OPAQUE
STATUS:CONFIRMED
SEQUENCE:0
LOCATION:https://meet.google.com/sqj-qimt-yno?hs=122&authuser=0
X-MICROSOFT-CDO-BUSYSTATUS:BUSY
BEGIN:VALARM
TRIGGER:-PT30M
ACTION:DISPLAY
DESCRIPTION:Reminder
END:VALARM
END:VEVENT
END:VCALENDAR