BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//TU/e ME//Events//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:Europe/Amsterdam
BEGIN:STANDARD
DTSTART:19701025T030000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:19700329T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:event-74@tuemeche.nl
DTSTAMP:20261007T234656Z
DTSTART;TZID=Europe/Amsterdam:20270119T110000
DTEND;TZID=Europe/Amsterdam:20270119T123000
SUMMARY:Entropy Regularization for Control and Estimation
DESCRIPTION:Speaker: Menno van Zutphen\nHost: Duarte Guerreiro Tomé Antu
 nes\n\n```Control and estimation for stochastic dynamical systems concern
  the synthesis of input policies and the reconstruction of state informat
 ion from measurements. Entropy provides a quantitative description of the
  randomness present in system trajectories\, disturbance models\, and bel
 ief distributions. The main contribution of this thesis is the developmen
 t of entropy-regularized methods for stochastic control and estimation. T
 his involves topics such as dynamic programming\, formal abstractions\, a
 nd recursive Bayesian filtering.\n\nThe central contribution concerns ent
 ropy-regularized control of continuous-state stochastic systems through f
 inite abstractions. Existing abstraction methods enable formal controller
  synthesis for objectives such as cumulative costs and temporal-logic spe
 cifications. Entropy-based trajectory objectives do not transfer through 
 these abstractions directly\, as discretization changes the entropy of th
 e induced trajectory distribution. This thesis derives bounds relating th
 e Kullback-Leibler (KL) divergence to uniform of a continuous trajectory 
 distribution to that of its finite discretization. These bounds enable fo
 rmal entropy-aware controller synthesis for continuous-state systems\, tr
 ading cumulative cost against trajectory predictability while retaining g
 uarantees for the original system.\n\nA second contribution concerns robu
 st stochastic control. It generalizes KL-regularized robust-control formu
 lations by allowing adversarial disturbance distributions to be regulariz
 ed through both cross entropy with respect to an empirical model and the 
 entropy of the adversary itself. The resulting dynamic-programming recurs
 ion gives rise to the minsoftmax algorithm and places minimax\, stochasti
 c\, KL-regularized\, and H-infinity type viewpoints in one parameterized 
 formulation.\n\nA third contribution concerns Bayesian filtering. The tem
 pered Bayes filter modifies the recursive Bayesian update by tempering th
 e distributions that define the posterior. This yields a computationally 
 efficient modification of the Bayes filter that can improve predictive pe
 rformance under model mismatch. Specializing the construction to the line
 ar Gaussian setting yields the tempered Kalman filter.\n\nTwo further pap
 ers are included as supporting material. The entropy-regularized interval
  Markov decision process work supports the continuous-state abstraction c
 hapter. The optimal stopping work is included as a thesis appendix outsid
 e the main entropy-regularization arc.```\n\nMore info: https://www.tue.n
 l/en/research/researchers/menno-van-zutphen
LOCATION:Atlas 0.710
URL:https://tuemeche.nl/peoplepages/event.php?id=74
CATEGORIES:PhD Defense
END:VEVENT
END:VCALENDAR