BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:From learning differential operators to learning algorithms - Houm
 an Owhadi (CALTECH (California Institute of Technology))
DTSTART:20250715T093000Z
DTEND:20250715T103000Z
UID:TALK232576@talks.cam.ac.uk
DESCRIPTION:Most scientific and engineering challenges can be organized al
 ong a complexity ladder.\nRight above interpolation lies the learning of d
 ifferential operators and their solution operators\, an area where Gaussia
 n Process/Kernel methods\, come with rigorous guarantees and achieve SOTA 
 in terms of data-efficiency and robustness. This talk then ascends to the 
 ladder&rsquo\;s current frontier: algorithm synthesis. Here we introduce a
  computational‑language&ndash\;processing framework that tokenizes low
 ‑level computational actions and uses an ensemble‑based Monte‑Carlo 
 Tree Search combined with reinforcement learning to assemble algorithms ta
 ilored to individual problem instances.&nbsp\; We conclude by discussing w
 here this ladder is taking us.\nThe first part of this talk is a joint wor
 k with based on joint work with Yasamin Jalalian\, Juan Felipe Osorio Rami
 rez\, Alexander Hsu\, and Bamdad Hosseini.\nThe second part is joint work 
 with Theo Bourdais\, Abeynaya Gnanasekaran and Tuhin Sahai.\n&nbsp\;
LOCATION:Seminar Room 2\, Newton Institute
END:VEVENT
END:VCALENDAR