Charles Margossian
@charlesm993
Research fellow @FlatironCCM and @mcmc_stan developer. Interested in computational and applied statistics. Opinions are my own.
ID: 1453355223116828678
http://charlesm93.github.io 27-10-2021 13:38:38
176 Tweet
575 Followers
169 Following
📃 My latest paper "Stochastic Localization via Iterative Posterior Sampling" (arxiv.org/abs/2402.10758) has been accepted at #ICML2024. 👏 It was a real team work between me and my co-authors Maxence Noble, Marylou Gabrié (Marylou Gabrié) and Alain Durmus.
Just arrived at Venice for the ISBA World Meeting. I will be presenting my work on sampling with stochastic localization during the "Monte Carlo algorithms for modern hardware" invited session chaired by Charles Margossian Charles Margossian (Thursday 4:30pm, Room 6A). See you there!
Reconnected with two former mentees ISBA: Emmanuel Mokel (left) and Stanislas Du Ché (right), who have done stellar work on MCMC using massive parallelization during their summer internships. Congrats Emmanuel Mokel on this year's best poster award!!
Training LLMs involves learning associations. We study training dynamics in a simple model that yields useful insight on the role of token interference and imbalance. Go talk to Vivien Cabannes and Berfin Simsek @ ICML at our #ICML poster #1114 tmrw 11:30! Paper: arxiv.org/abs/2402.18724
It was a fun (and difficult) exercise proposed by neptune.ai to summarize our paper in 100 seconds without any rehearsal. I liked it! Thread about the paper: x.com/oharub/status/…
Details from Aki Vehtari - @[email protected] on the upcoming StanCon 2024 model selection tutorial: discourse.mc-stan.org/t/stancon-2024…
Big thanks to Alexandre Andorra & Chris Wymant @chriswymant.bsky.social for a great #StanCon 2024 panel! We covered epidemiology vs biology, infectious diseases vs NCDs, generative emulators, graph GPs, sequential data collection, and, of course, #ProbabilisticProgramming.