BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.2.3.1//EN TZID:America/New_York X-WR-TIMEZONE:America/New_York BEGIN:VEVENT UID:849@fds.yale.edu DTSTART;TZID=America/New_York:20250324T153000 DTEND;TZID=America/New_York:20250324T170000 DTSTAMP:20250916T142148Z URL:https://fds.yale.edu/events/sds-colloquium-thuy-duong-june-vuong-mille r-institute-berkeley-efficiently-learning-and-sampling-from-multimodal-dis tributions-using-data-based-initialization/ SUMMARY:S&\;DS Colloquium: Thuy-Duong "June" Vuong (Miller Institute\, B erkeley)\, "Efficiently learning and sampling from multimodal distribution s using data-based initialization" DESCRIPTION:\nAbstract: Learning to sample is a central task in generative AI: the goal is to generate (infinitely many more) samples from a target distribution $\\mu$ given a small number of samples from $\\mu.$ It is wel l-known that traditional algorithms such as Glauber or Langevin dynamics a re highly inefficient when the target distribution is multimodal\, as they take exponential time to converge from a \\emph{worst case start}\, while recently proposed algorithms such as denoising diffusion (DDPM) require i nformation that is computationally hard to learn. In this talk\, we propos e a novel and conceptually simple algorithmic framework to learn multimoda l target distributions by initializing traditional sampling algorithms at the empirical distribution. As applications\, we show new results for two representative distribution families: Gaussian mixtures and Ising models. When the target distribution $\\mu$ is a mixture of $k$ well-conditioned G aussians\, we show that the (continuous) Langevin dynamics initialized fro m the empirical distribution over $\\tilde{O}(k/\\epsilon^2)$ samples\, wi th high probability over the samples\, converge to $\\mu” in $\\tilde{O} (1)$-time\; both the number of samples and convergence time are optimal. W hen $\\mu$ is a low-complexity Ising model\, we show a similar result for the Glauber dynamics with approximate marginals learned via pseudolikelih ood estimation\, demonstrating for the first time that such low-complexity Ising models can be efficiently learned from samples.”\n\n\n\nBased on joint work with Frederic Koehler and Holden Lee.\n\n\n\nSpeaker bio: I am a postdoctoral fellow at the Miller Institute\, Berkeley\, hosted by Alis tair Sinclair.I have a broad interest in theoretical computer science. My current research interest is in algorithms for sampling from complex high- dimensional distributions\, with applications to statistical physics\, gen erative AIs\, and other fields.I received my PhD student in Computer Scien ce at Stanford University in 2024\, advised by Nima Anari and Moses Cha rikar. My PhD was partially supported by a Microsoft Research PhD Fellows hip (2022-2024).I received my Bachelor's degrees in Mathematics and Compu ter Science at the Massachusetts Institute of Technology (MIT) in 2019.I a m joining UC San Diego CSE as an assistant professor in January 2026.\n\n\ n\nhttps://thuyduongvuong.github.io\n CATEGORIES:FDS Events,Statistics & Data Science Seminar LOCATION:Yale Institute for Foundations of Data Science\, Kline Tower 13th Floor\, Room 1327\, New Haven\, CT\, 06511\, United States X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Kline Tower 13th Floor\, Ro om 1327\, New Haven\, CT\, 06511\, United States;X-APPLE-RADIUS=100;X-TITL E=Yale Institute for Foundations of Data Science:geo:0,0 END:VEVENT BEGIN:VTIMEZONE TZID:America/New_York X-LIC-LOCATION:America/New_York BEGIN:DAYLIGHT DTSTART:20250309T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 TZNAME:EDT END:DAYLIGHT END:VTIMEZONE END:VCALENDAR