Associate Professor & EPSRC Fellow @ UCL. Post-Bayesian seminar series sign-up @ https://xitter.cf/t.co/a0MyAQOh16 Research mission @ https://xitter.cf/t.co/kNIjvCrGne
London
Joined April 2019
- Tweets397
- Following501
- Followers1.5K
- Likes1K
Jeremias Knoblauch retweeted
I'm on the academic job market!
I design and analyze probabilistic machine-learning methods---motivated by real-world scientific constraints, and developed in collaboration with scientists in biology, chemistry, and physics.
A few highlights of my research areas are:
ALT Probabilistic machine learning in scientific pipelines.
Jeremias Knoblauch retweeted
I am pleased to share our new NeurIPS paper for online Bayesian inference for neural networks. Instead of focusing on updating the parameter posterior, we work with the predictive posterior (which makes much more sense for non-identifiable models, and gives us more algorithmic freedom for developing faster methods).
Our paper “Martingale Posterior Neural Networks for Fast Sequential Decision Making” has been accepted at #neurips2025!
Joint work with @l_sbetancourt, @AlvaroCartea and @sirbayes
Blog: grdm.io/posts/bnn-without-ba…
Paper: arxiv.org/abs/2506.11898
Code: github.com/gerdm/martingale-…
Jeremias Knoblauch retweeted
Today is the day of the pre-ICML event at UCL! Come check out the exciting work from academics, industry researchers, postdocs and PhD students from around London: sites.google.com/view/pre-ic… @stats_UCL @uclcsml
Jeremias Knoblauch retweeted
Videos from the recent UCL workshop on ‘Advances in Post-Bayesian methods’ are now available on YouTube!
Check these out if you couldn’t make it: tinyurl.com/AdvPostBayes
@LauchLab @matialtamiranom
Jeremias Knoblauch retweeted
There’s been lots of interest in gradient flows, but I’ve been a bit sceptical as I wasn’t clear on the advantages.
This paper shows that MMD gradient flows can integrate a large class of functions exactly without needing to converge to a global minimum: stationarity is enough!
New paper on Stationary MMD points 📣
arxiv.org/pdf/2505.20754
1️⃣ Samples generated by MMD flow exhibit 'super-convergence'
2️⃣ A discrete-time finite-particle convergence result for MMD flow
Joint work with Toni Karvonen, Heishiro Kanagawa, @fx_briol, Chris J. Oates
Jeremias Knoblauch retweeted
🆕 The ELLIS Unit London (@ucl) will host the first workshop uniting disparate subfields of post-Bayesian methods, including PAC Bayes, generalized Bayes, predictive resampling, Martingale posteriors, and online learning.
📍 London 🏴
📅 May 15-16
🔗 bit.ly/3GgwZ0G
Jeremias Knoblauch retweeted
Don't miss the very first workshop on post-Bayesian methods at UCL on May 15-16. Registration is now open!
We are excited to announce that registration for the inaugural post-Bayes workshop on May 15./16. at UCL is now open!
Website: postbayes.github.io/workshop…
Registration link: tinyurl.com/postBayesWorksho…
Jeremias Knoblauch retweeted
Looking forward to speaking at the Post-Bayes workshop in May!
We are excited to announce that registration for the inaugural post-Bayes workshop on May 15./16. at UCL is now open!
Website: postbayes.github.io/workshop…
Registration link: tinyurl.com/postBayesWorksho…
Jeremias Knoblauch retweeted
Why posteriors MUST be modified, and a simple technic for setting them up.
Dr. French's Modified Posteriors (1937) @ISBA_events
abebooks.com/Dr-Frenchs-Modi…
Jeremias Knoblauch retweeted
Congratulations to @UCL / @stats_UCL's @matialtamiranom on being one of the 2024-2025 @Bloomberg #DataScience Ph.D. Fellows!
Learn more about Matías’ research focus and our latest cohort of Ph.D. Fellows: bloom.bg/4itDDiN
#AI #ML #NLProc #NeurIPS2024
ALT Matías Altamirano of the Department of Statistical Science at University College London (UCL) is a 2024-2025 Bloomberg Data Science Ph.D. Fellow.
Jeremias Knoblauch retweeted
Super happy and grateful to be awarded the Bloomberg fellowship. Huge thanks to my amazing supervisors @LauchLab and @fx_briol , as well as all my collaborators for their support!
Congratulations to @UCL / @stats_UCL's @matialtamiranom on being one of the 2024-2025 @Bloomberg #DataScience Ph.D. Fellows!
Learn more about Matías’ research focus and our latest cohort of Ph.D. Fellows: bloom.bg/4itDDiN
#AI #ML #NLProc #NeurIPS2024
ALT Matías Altamirano of the Department of Statistical Science at University College London (UCL) is a 2024-2025 Bloomberg Data Science Ph.D. Fellow.
Jeremias Knoblauch retweeted
Interested in reducing the computational cost of running simulations in simulation-based inference (SBI)?
Check out our new “Cost-aware SBI” paper w. @DaolangHuang @samikaski @fx_briol
arxiv.org/abs/2410.07930
Thread👇
Jeremias Knoblauch retweeted
Applications are invited for 1 PhD Studentship (UK fee only) based at the Department of Statistical Science, University College London. The position can involve any topic within the Statistical Science remit.
Deadline for applications: 16 August 2024
ucl.ac.uk/statistics/prospec…
Jeremias Knoblauch retweeted
@matialtamiranom will present two papers at @icmlconf on how generalised Bayes can make probabilistic ML robust and scalable. Say hi to him!
Spotlight paper on GPs -Wed HallC 4-9 #1506: arxiv.org/abs/2311.00463
Paper on KF -Tues HallC 4-9 #1602: arxiv.org/abs/2405.05646
@LauchLab
Jeremias Knoblauch retweeted
Happy to share that our paper "Robust and Conjugate Gaussian Process Regression" has been accepted as a spotlight at #ICML2024. We proposed a robust Gaussian process regression at virtually no added cost! Joint work with @fx_briol and @LauchLab, all research @stats_UCL. [1/6]
Jeremias Knoblauch retweeted
Happy to share that our outlier-robust Kalman filter (WoLF 🐺) has been accepted to #ICML!
arxiv.org/abs/2405.05646
We propose a novel filtering algorithm that is provably robust and has similar computation cost to the KF.
Now, why care about robust filtering?
Jeremias Knoblauch retweeted
We have a new webpage for the "fundamentals of statistical machine learning"research group at @stats_UCL: fsml-ucl.github.io. Check it out to find out what we have been up to!
@LauchLab @matialtamiranom @MNaslidnyk @IevaKazlauskai @ucl
Jeremias Knoblauch retweeted
📢 I am looking for 2 postdocs in my ERC project NEMESIS (tinyurl.com/4d2ry2vj) to develop advanced statistical models for complex (criminal 🕵️) networks‼️
Deadline: April 30, 2024
Duration: 1 year, renewable
Where: BIDSA @Unibocconi
Info-Application: jobmarket.unibocconi.eu/?id=…
Jeremias Knoblauch retweeted
#ContinualLearning is NP-hard.
@LauchLab, @syimplectic, and @tommy_da_cat brilliantly proved it in their ICML 2020 paper (arxiv.org/abs/2006.05188) using set theory.
This provoked the following thought process that eventually made me depart from conventional CL. 🧵(1/n)
