BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20250309T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20241103T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20241110T210114Z UID:EA833D62-BE99-4AC5-A101-718D9CCDDACF DTSTART;TZID=America/New_York:20241106T190000 DTEND;TZID=America/New_York:20241107T200000 DESCRIPTION:Abstract: Many constrained sequential decision-making processes \, such as safe AV navigation\, wireless network control\, caching\, cloud computing\, etc.\, can be cast as Constrained Markov Decision Processes ( CMDP). Reinforcement Learning (RL) algorithms have been used to learn opti mal policies for unknown unconstrained MDP. Extending these RL algorithms to unknown CMDP brings the additional challenge of maximizing the reward a nd satisfying the constraints. In this talk\, I will present algorithms th at can learn safe policies effectively.\n\nIn the second part of the talk\ , I will demonstrate how the theoretical understanding of the constrained MDP can help us to develop algorithms for practical applications. As an ap plication\, I show how to learn to obtain optimal beam directions under ti me-varying interference-constrained channels for a mobile service robot. O ptimal beam selection in mmWave is challenging because of its time-varying nature. We propose a primal-dual Gaussian process bandit with adaptive re initialization to handle non-stationarity and interference constraints. We demonstrate how our approach learns to adapt effectively to time-varying channel conditions.\n\nCo-sponsored by: IEEE North Jersey Section \n\nSpea ker(s): Dr. Arnob Ghosh\, \n\nAgenda: \nNov 6th\,\n\nTalk: 7:00 PM - 8:00 PM\n\nDiscussion Q/A: 8:00 PM - 8:15 PM\n\nRoom: ECE 202\, Bldg: Electrica l and Computer Engineering\, 154 Summit Street\, Newark\, New Jersey\, Uni ted States\, 07102 LOCATION:Room: ECE 202\, Bldg: Electrical and Computer Engineering\, 154 Su mmit Street\, Newark\, New Jersey\, United States\, 07102 ORGANIZER:anisha_apte@ieee.org SEQUENCE:11 SUMMARY:Learn to Solve Constrained Markov Decision Process Efficiently URL;VALUE=URI:https://events.vtools.ieee.org/m/441503 X-ALT-DESC:Description: <br /><p class="MsoNormal"><strong>Abstract:</stron g> Many constrained sequential decision-making processes\, such as safe AV navigation\, wireless network control\, caching\, cloud computing\, etc.\ , can be cast as Constrained Markov Decision Processes (CMDP). Reinforceme nt Learning (RL) algorithms have been used to learn optimal policies for u nknown unconstrained MDP. Extending these RL algorithms to unknown CMDP br ings the additional challenge of maximizing the reward and satisfying the constraints. In this talk\, I will present algorithms that can learn safe policies effectively.&nbsp\;</p>\n<p class="MsoNormal">In the second part of the talk\, I will demonstrate how the theoretical understanding of the constrained MDP can help us to develop algorithms for practical applicatio ns. As an application\, I show how to learn to obtain optimal beam directi ons under time-varying interference-constrained channels for a mobile serv ice robot. Optimal beam selection in mmWave is challenging because of its time-varying nature. We propose a primal-dual Gaussian process bandit with adaptive reinitialization to handle non-stationarity and interference con straints. We demonstrate how our approach learns to adapt effectively to t ime-varying channel conditions.</p>\n<p>&nbsp\;</p><br /><br />Agenda: <br /><p>Nov 6th\,</p>\n<p>Talk: 7:00 PM - 8:00 PM</p>\n<p>Discussion Q/A: 8: 00 PM - 8:15 PM</p>\n<p>&nbsp\;</p>\n<p>&nbsp\;</p> END:VEVENT END:VCALENDAR