@thdbui

Senior Lecturer (US equiv Assoc Prof) in Machine Learning @ANUComputing. Previously @Sydney_Uni, @Uber AI and @CambridgeMLG

thangbui.github.io/
Joined June 2020
  • Tweets36
  • Following222
  • Followers589
I'm so happy to share that I’ll be joining @UofT as an Assistant Professor of Statistical Sciences and Computer Science, with an appointment at the @VectorInst, in 2026! I'm recruiting postdocs and PhD students: timrudner.com! Please help me spread the word! 🧵(1/5)
26
73
15
372
41,914
Thang Bui retweeted
lexing xie@lexing11 Dec 2024
Announcing 2x Research Fellows positions on NLP and machine learning, working w wonderful Jing Jiang, Thang Bui and myself + partners in Data61 - apply by 31 Dec or help spread the words jobs.anu.edu.au/jobs/researc… more detail in thread
1
2
4
1,083
Thank you @vincefort @markvanderwilk @EmtiyazKhan for organising this great @dagstuhl workshop! dagstuhl.de/24461
7
3
45
8,518
Check out this outstanding work from some of the best people I know, fellow @universe_tbd ers and the best co-convenor in the Universe! Great stuff, @jackm2003 @oneill_c @thdbui. Keep grokking!
Jack Miller
@jackm200327 Oct 2023
We have recently written a new paper on grokking: arxiv.org/abs/2310.17247. We show that the phenomenon is not limited to neural networks, this motivates a new hypothesis which seeks to explain it. See the thread below for more details. @thdbui and @oneill_c.

Grokking Beyond Neural Networks: An Empirical Exploration with...

In some settings neural networks exhibit a phenomenon known as \textit{grokking}, where they achieve perfect or near-perfect accuracy on the validation set long after the same performance has been...

arxiv.org
2
5
1,044
Absolutely thrilled to announce 🚀 AstroLLaMa from @universe_tbd! Introducing our specialized Llama-2 models tailored for astronomers 🌌. This is just the beginning! But our tests already show extraordinary text generation and embeddings for astronomy!💫 arxiv.org/abs/2309.06126
1
17
7
53
12,043
Happy to announce the release of a new language model, AstroLLaMA. Joint work with the amazing @TingAstro, @errai34 and many amazing collaborators @universe_tbd. arxiv.org/abs/2309.06126 1/n

AstroLLaMA: Towards Specialized Foundation Models in Astronomy

Large language models excel in many human-language tasks but often falter in highly specialized domains like scholarly astronomy. To bridge this gap, we introduce AstroLLaMA, a 7-billion-parameter...

arxiv.org
1
17
1
46
4,418
Thang Bui retweeted
Yee Whye Teh@yeewhye3 Aug 2022
Are you an ML research engineer interested in principled approaches to deep learning and/or Bayesian learning? Are you interested in joining a dynamic, kind and friendly team working on cutting edge research? If so, please apply @DeepMind London office! boards.greenhouse.io/deepmin…
61
2
365
Come say hi the "Deep Learning through Information Geometry" workshop today! We're presenting our work on q-Paths, a generalization of the geometric mixture path commonly used in the AIS, TVO, PT, etc.@brekelmaniac @thdbui @frankdonaldwood @aram_galstyan @gesteller @FrnkNlsn
1
2
10
In "AIS with q-Paths", we extend beyond the geometric mixture & exponential family to the generalized power mean & Tsallis's q-exponential family, and use this for alternative paths in mcmc sampling! talk: slideslive.com/38942377/anne… paper: openreview.net/forum?id=ZBJ2…

Rob Brekelmans, Vaden Masrani, Thang Bui, Frank Wood, Aram Galstyan, Greg Ver Steeg, Frank Nielsen...

slideslive.com
1
4
5
So our paper "Hierarchical Gaussian Process Priors for Bayesian Neural Network Weights" with @thdbui is up today at #NeurIPS2020 poster session 7 at 21:00 PST. Let me give a quick recap of the innovations in this work. 1/n
2
20
1
146
Load more