←
Adrián Javaloy1,200 posts
Adrián Javaloy avatar

Adrián Javaloy

@javaloyML

Lecturer in AI and ML at the University of Bath. Previously @InfAtEd @SIC_Saar @MPI_IS @UMU.

◷ Joined February 2020▧ 80 Photos and videos
Adrián Javaloy avatar
Adrián Javaloy@javaloyML·9 Jun 2023

Hi! I am happy to announce our latest work (with @ML_pablosm and @IValeraM) ... 🥁 🚨 Causal normalizing flows: from theory to practice 🚨 I am really excited about this work! Check this image, and scroll down if you want to know more 👀 1/10

Post attachment
Adrián Javaloy avatar
Adrián Javaloy@javaloyML·Nov 16

🎓 Causal generative models are dear to my scientific heart and I am certain there is much more to say, so if you are a curious mind with a taste for excellence and human-written research, join me in beautiful Bath to start your scientific career! findaphd.com/phds/project/ca…

Adrián Javaloy avatar
Adrián Javaloy@javaloyML·Nov 16

🤔 Would have I changed the world had I accepted that PhD offer? 🤓 If counterfactual questions like these don't let you sleep at night, I am looking for a PhD student motivated to push with me the boundaries of what is possible with causality and GenAI

Thomas Kipf avatar
⟳ Adrián Javaloy retweeted
Thomas Kipf@tkipf·May 13

Confession: I never had a single work-related sleepless night or ever pulled an all-nighter during my career incl. PhD. Don’t sacrifice your health. Sleep is a superpower — your brain on 8hrs of sleep is a lot smarter than your brain on sleep deprivation. Don’t listen to people who tell you to chronically sacrifice sleep for work. Sacrificing sleep for your kids/family is a different story.

I doubt all those things are really possible. Infact I believe, you are not doing a good PhD unless you have sleepless nights. Definitely just working on your thesis is possible if you follow a 9-6 schedule, but a good PhD which involves exploring, colabs, etc needs extra hours

Adrián Javaloy avatar
Adrián Javaloy@javaloyML·Feb 12

More importantly, we show that this comes at no cost in performance, and we can even train non-structured-decomposable squared circuits* (* That is, circuits which cannot be efficiently squared)

Post attachment
Adrián Javaloy avatar
Adrián Javaloy@javaloyML·Feb 12

As a result, we can train really large squared circuits while saving both time and memory! At 357M parameters, we (⟂) use: - 12 GiB vs 18 GiB (33% reduction!) - 0.29ms vs 0.52ms per iteration (44% faster!)

Post attachment
Adrián Javaloy avatar
Adrián Javaloy@javaloyML·Feb 12

Yes, we can! 💡 We generalize both ideas and propose to use orthogonality constraints to parametrize *already normalized* squared circuits That way, we completely avoid squaring them during training!

Adrián Javaloy avatar
Adrián Javaloy@javaloyML·Feb 12

In the tensor network community, this issue can be avoided for specific cases using canonical forms And in the circuit community, determinism (i.e. non-overlapping supports) makes the square tractable, although it is too restrictive... 🤔 Can we go expand on these ideas?

Adrián Javaloy avatar
Adrián Javaloy@javaloyML·Feb 12

One way of increasing the expressiveness of probabilistic circuits is to square them (multiply a circuit with itself). 😔 However, this imposes a quadratic cost in the circuit size, as we need to re-normalize it to ensure that it encodes a valid probability.

Adrián Javaloy avatar
Adrián Javaloy@javaloyML·Feb 12

I am a bit late to the party, but I am happy to share that our latest work was accepted to #ICLR2026 🥳🥳 📜 How to Square Tensor Networks and Circuits Without Squaring Them arxiv.org/abs/2512.17090

antonio vergari ⚔️ avatar
⟳ Adrián Javaloy retweeted
antonio vergari ⚔️@tetraduzione·11 Nov 2024

Less than two weeks to submit your papers on: 📈 #lowrank adapters and #factorizations 🧊 #tensor networks 🔌 probabilistic #circuits 🎓 #theory of factorizations to the first workshop on connecting them in #AI #ML at @RealAAAI please share! 🔁 👇👇👇 april-tools.github.io/colora…

many of the recent successes in #AI #ML are due to #structured low-rank representations! but...What's the connection between #lowrank adapters, #tensor networks, #polynomials and #circuits? join our @RealAAAI workshop to know the answer! 👇👇👇 april-tools.github.io/colora…

Post attachment
@𝘾𝙝𝙖𝙏𝙤@tech.lgbt in Mastodon avatar
⟳ Adrián Javaloy retweeted
@𝘾𝙝𝙖𝙏𝙤@tech.lgbt in Mastodon@ChaToX·5 Nov 2024

I am recruiting a PhD student to being in September 2025, through the ELLIS call for PhD students: ellis.eu/news/ellis-phd-prog… (deadline Nov 15th)