AutoML.org

research groups on Automated Machine Learning

[AutoML.org @AutoML_org]

AutoML in the weights of a neural network.

[Samuel Müller @SamuelMullr]

This might be the first time after 10 years that boosted trees are not the best default choice when working with data in tables.
Instead a pre-trained neural network is, the new TabPFN, as we just published in Nature 🎉

[Frank Hutter @FrankRHutter]

The data science revolution is getting closer. TabPFN v2 is published in Nature: https://t.co/Ybb15pnZ5P On tabular classification with up to 10k data points & 500 features, in 2.8s TabPFN on average outperforms all other methods, even when tuning them for up to 4 hours🧵1/19

[Sathya Kamesh @SathyaKamesh98]

We are elated to introduce our most recent work on time-series foundation models - Mamba4Cast: Efficient Zero-Shot Time Series Forecasting with State Space Models.
Authors: @o_swelam, Sathya Kamesh, @julien_siems, David Salinas, @FrankRHutter

Link: https://t.co/1JkVRqm7VS

[Günter Klambauer @gklambauer]

Unlocking State-Tracking in Linear RNNs Through Negative Eigenvalues

Forget my earlier post, this is the cool one! :)

Analysis of STATE TRACKING capabilities of "linear RNNs" (GLA, MAMBA, mLSTM).

P: https://t.co/xdcOnUrKjE

[AutoML.org @AutoML_org]

Excited to share our work on this simple yet powerful method for linear RNNs like Mamba or DeltaNet to track states without increasing computational complexity. From @PontilGroup and @FrankRHutter's group

[Riccardo Grazzi @riccardograzzi]

LLMs can now track states, finally matching this cat!
And we prove it.

But how? 🧵👇

1/ Paper: https://t.co/aKvrqYtkWh
with @julien_siems @jkhfranke @ZelaArber  @FrankRHutter   @MPontil

[AutoML.org @AutoML_org]

This work was done jointly by @JakeMRobertson under the supervision of the wonderful @NoorAwad and @FrankRHutter.

Link to paper: https://t.co/VFZwPQGdYs
Blog post: https://t.co/7SmTwse0Jq

#AI #FairnessInAI #MachineLearning #EthicsInAI #ACM #ManyFairHPO

[AutoML.org @AutoML_org]

When it comes to fairness in AI, "fair" can mean different things to different people. We saw this challenge as an opportunity to innovate!

🌟 Introducing ManyFairHPO: a human-in-the-loop optimization framework!

[AutoML.org @AutoML_org]

ManyFairHPO helps practitioners navigate fairness metric trade-offs, assess conflicts, and make context-aware model selections amidst nuanced fairness dilemmas.

Join us for an engaging discussion on October 22nd at the AAAI/ACM AI Ethics and Society Conference in San Jose!

[Samuel Müller @SamuelMullr]

Transformers perform remarkable generalizations in the in-context learning setting.
E.g. when trained only on step functions, the model generalizes to smooth predictions when given a smooth input.
(1/n, a paper thread)

[CarolaDoerr @CarolaDoerr19]

What a great pleasure it was to host the @automl_conf here in Paris this week.
Big shoutout to all co-organizers and in particular to the amazing @ElenaRaponi_ @anjajankovic Simon Provost and to the online chairs Gabi Kadlecová and @AndreBiedenkapp
See you in NYC next year 😃

[SLDS / Bernd Bischl @BBischl]

We have another opening for a PhD in AutoML @LMU_Muenchen. Apply now at https://t.co/3YJrnAAlil
#PhD #MachineLearning #Statistics #DataScience #AutoML

[AutoML Conference @automl_conf]

In less than a month, the AutoML Conference 2024 will be in Paris. I don't think that we can quite compete with the #Olympics, but we will give our best ;-)

[AutoML.org @AutoML_org]

Let's make ML fairer for everyone! 🤖⚖️
Link to paper: https://t.co/5xRM00P3bN

@JakeMRobertson @noahholl @__NoorAwad__ @FrankRHutter

#AI #Fairness #CausalInference #Transformers #MLResearch

[AutoML.org @AutoML_org]

🌟 Excited to share our latest work on counterfactual fairness in #MachineLearning at the ICML Workshop on Next Gen. AI Safety in Vienna 🌟

We introduce FairPFN, a transformer trained on synthetic data to remove gender, age, and racial bias in -sensitive real-world ML problems.

[AutoML.org @AutoML_org]

Our experiments show some promising results, paving the way transformers in causal and counterfactual fairness with exciting applications to law, healthcare, and finance, as well as extensions to fairness pre-processing, path-specific effects, and multi-objective optimization.

[Neeratyoy Mallik @neeratyoy]

In #Vienna for #ICML2024 to present our latest work!
With @_herilalaina_, @StAdriaensen, @karibbov, @Edberg_Wardman, and @FrankRHutter, we introduce,

ifBO: In-Context Freeze-Thaw Bayesian Optimization for Hyperparameter Optimization.

[Matthias Feurer @__mfeurer__]

Joint work with @LindauerMarius, Florian Karl, Julia Moosbauer, @ATornede, @amuellerml, @FrankRHutter and @BBischl

[Matthias Feurer @__mfeurer__]

📝In the meantime you can also read a short summary blog post at https://t.co/KJfFpYBQMF or watch a video summary at https://t.co/qXkyPPcql2 🎥 2/3

[Matthias Feurer @__mfeurer__]

Wondering how humans should be involved in designing #AutoML solutions 🤔? Check out our #ICML2024 paper: "Position: A Call to Action for a Human-Centered AutoML Paradigm"! 📄✨ https://t.co/TmB1p7HIhw

Drop by at our poster on Thu, Jul 25 at 11:30 AM in Hall C 4-9 #2003 📅

1/3