It looks like I have to add Wiley to the list of publishers I refuse to publish with or review for 🤷🏻‍♂️

We are hiring. We have a post-doc position on machine learning for large-scale shared representations of dynamic environments with a focus on developing novel neural distributed compression algorithms. Please spread the news, opening here: www.imperial.ac.uk/jobs/search-...

184 years earlier than predicted, London has met the legal limits for toxic NO2 for the first time. This historic milestone proves that bold action like expanding ULEZ works, protecting children’s health and giving millions cleaner air to breathe.

very happy that our paper on memorization in diffusion models: "Tracing the Roots: Leveraging Temporal Dynamics in Diffusion Trajectories for Origin Attribution" has been accepted at #NeurIPS2025

I suppose this is the point when I should say again that: the amount of money the videogames industry brings into the UK economy every year is *more than double* the value of the fishing and steel industries *combined* what this country is really good at is producing culture

If you are at #ICML2025 come to our spotlight poster "LotteryCodec: Searching the Implicit Representation in a Random Network for Low-Complexity Image Compression" on Thursday: eedavidwu.github.io/LotteryCodec/

as someone only able to use the left hand, I'm really looking forward to attending this concert (is it too late for me to become a great pianist like Nicholas McCarthy? 😏): www.theguardian.com/music/2025/j...

new preprint on diffusion posterior sampling using invertible neural networks is available here: arxiv.org/abs/2505.12935

Ingrid Daubechies' Bakerian lecture is now available here: www.youtube.com/watch?v=nk0Y...

There's a lot of big news today/this week, but Apple just dropped a nuke of a paper about LLMs & LRMs, specially around "high-complexity tasks where both models experience complete collapse." This is the biggest sign yet that if AI ever lives up the hype, it won't be via those approaches:

I am very happy that The Bakerian Medal and Lecture 2025 is awarded to Ingrid Daubechies and I am looking forward to attending her lecture tomorrow: royalsociety.org/medals-and-p...

I thoroughly enjoyed the EEE Alumni Event organized by @imperialeee.bsky.social It’s inspiring to hear successful young professionals with a PhD with us speak openly also about their personal failures and rejections. When such honesty comes from accomplished individuals, it feels even more powerful.

I have cleaned up the notebooks for my course on Optimal Transport for Machine Learners and added links to the slides and lecture notes. github.com/gpeyre/ot4ml

This is a good article. It's so easy to generate results with AI tools that we are rushing to publish them without doing the hard work of checking that they really hold up. This problem predates AI though. So much computational modelling has the same problems www.understandingai.org/p/i-got-fool...

I have cleaned a bit my lecture notes on Optimal Transport for Machine Learners arxiv.org/abs/2505.06589

The 2025 call for the Imperial College Research Fellowships (ICRFs) is now open (www.imperial.ac.uk/research-and...). Any top researcher with less than four years of Post-doc experience and interested in computational imaging is welcome to contact me.

just back from the party to celebrate the reopening of the Sainsbury Wing at the National Gallery. The Gallery in the evening is as stunning as ever. The new wing will open to the public on Saturday.

Our research group in the department of Mathematics and Computer Science at the University of Basel (Switzerland) is looking for several PhD candidates and one post-doc who have a theoretical background in optimization and machine learning or practical experience in the field of reasoning.

When I have time, I enjoy walking around relatively unexplored areas of London. This is the Three Mills along "The Line Sculpture Trail"

It's not often that you get to spend two days discussing bio-imaging challenges with Nobel Prize winners. Hence, this photograph 😏

Inside arXiv—the Most Transformative Platform in All of Science https://www.wired.com/story/inside-arxiv-most-transformative-code-science/

Very happy to be participating in this Wellcome Trust workshop on barriers and challenges in bio-imaging. The team includes top scientists and Nobel prize winners! Learned a lot in day one and looking forward to day two.

Characterizing finely the decay of eigenvalues of kernel matrices: many people need it, but explicit references are hard to find. This blog post reviews amazing asymptotic results from Harold Widom (1963!) and proposes new non-asymptotic bounds. francisbach.com/spectrum-ker...

Our pre-print on image transmission using joint source-channel coding techniques and generative diffusion models is available here: arxiv.org/abs/2503.12484

very proud of our new coffee machine. @imperialeee.bsky.social we cannot turn coffee to gold, but we can certainly turn it to beer 😏

European alternatives for digital services. Sovereignty matters.

🚀 Fully funded PhD in Neurosymbolic AI for Finance 💰🤖 🔍 Work at Imperial College London on cutting-edge AI for financial applications! [1/3]

This is an extraordinary speech. That a Canadian leader is making a speech like this about the US is hard to get your head around. Does he only mention one US President by name (Kennedy)?

This review paper by @guillaume-garrigos.com on SGD-related algorithms is a fantastic resource, offering elegant, self-contained, and concise proofs in a single, accessible reference. arxiv.org/pdf/2301.11235

Our repo: github.com/open-thought... Open code, Open reasoning data (114k and growing), Open weight models. Please let us know if you want to participate in the Open Thoughts community effort. (2/n) www.openthoughts.ai

What if we had the data that DeepSeek-R1 was post-trained on? We announce Open Thoughts, an effort to create such open reasoning datasets. Using our data we trained Open Thinker 7B an open data model with performance very close to DeepSeekR1-7B distill. (1/n)

No way #MathSky

We just did a crazy 48h sprint to create the best public reasoning dataset using Berkeley's Sky-T1, Curator and DeepSeek R1. We can get o1-Preview reasoning on a 32B model and 48x less data than Deepseek. t.co/WO5UV2LZQM

My department @imperialeee.bsky.social @imperialcollegeldn.bsky.social is hiring for a new faculty member in "intelligent systems" that might be of interest to some of you working in AI/ML/etc. Deadline is Feb 12th. Can strongly recommend as a place to work and live. www.jobs.ac.uk/job/DLJ503/l...

We are hiring at assistant/associate professor level. We are particularly interested in applicants working on AI, Machine Learning (ML), and data sciences in topics relevant to the work of the department. Deadline 12th of February: www.imperial.ac.uk/jobs/search-....

Back in the days of punch cards, when dinosaurs roamed the Earth and a software patch meant a piece of tape...🦖

What do you find to be the best (and worst?) parts of the academic community moving over to Bluesky? I personally enjoy the more friendly community here and the lack of algorithm baiting, but do feel like it is a bit insular and could limit our reach to the general public. 🧪 #AcademicSky

We are pleased to welcome Professor Tim Green @otherprofgreen.bsky.social as the new Head of the Department of Electrical and Electronic Engineering. Tim has a long history at Imperial - he first joined our department as an undergraduate in 1982! Find out more: www.imperial.ac.uk/news/259633/...

So happy to see that Ingrid Daubechies will receive the National Medal of Science in 2025. She has been an inspirational figure and it has also been an honour to collaborate with her over the last few years on the technical study of Old Master paintings. today.duke.edu/2025/01/ingr...

Have to share that Ingrid Daubechies, one of my true mathematical heroes, was one of the honorees awarded the prestigous National Medal of Science by Pres. Biden. She is a role model to me! Congrats to Dr Daubechies!!! #MathSky today.duke.edu/2025/01/ingr...

I have been having fun turning the Mario Brothers into a 1940s industrial film using Veo 2.

On the 2nd I went ice skating at the Battersea Power Station ice rink. The area is now a tiny bit too commercial for my liking but the power station is as iconic as ever.

I have updated my course notes on Optimal Transport with a new Chapter 9 on Wasserstein flows. It includes 3 illustrative applications: training a 2-layer MLP, deep transformers, and flow-matching generative models. You can access it here: mathematical-tours.github.io/book-sources...

Enjoyed the "Van Gogh: Poets and Lovers" exhibition at the National Gallery. If you are in London I recommend you try to go there. I also have a special attachment to the "Sunflowers" of the NG because with the NG scientific department we performed a technical study of that painting.

Also a Python code to display trajectories github.com/mathematical...

I wrote a summary of the main ingredients of the neat proof by Hugo Lavenant that diffusion models do not generally define optimal transport. github.com/mathematical...

This podcast on "Radiance Fields and The Future" with @jonbarron.bsky.social is streaming right now! www.youtube.com/watch?v=h9vq...

Anne Gagneux, Ségolène Martin, @quentinbertrand.bsky.social Remi Emonet and I wrote a tutorial blog post on flow matching: dl.heeere.com/conditional-... with lots of illustrations and intuition! We got this idea after their cool work on improving Plug and Play with FM: arxiv.org/abs/2410.02423

Interesting post blogs.lse.ac.uk/impactofsoci...