Základní pravidlo sociálních sítí: NIKDY NEKOMENTUJ POUZE TITULEK NEBO CITÁT!! Pokud si ODBORNÝ rozhovor neposlechneš celý, nekomentuj ho! Samozřejmě to občas porušuju také, ale tento rozhovor s panem Mikolovem opravdu NENÍ útokem na "studium žab v Amazonii", ale na diletantský

Reposting to increase the proba that it gets in the right X bubble!

This week @Gabuinml will talk about her latest work on performance prediction for neural architecture search!Thursday 1pm CET

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 😃

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 ;-)

As a bonus, we showed that some zero-cost proxies are not as great as we thought 🙂 Excited to discuss the work in Vienna! 4/4

In Neural architecture search, performance predictors can be quite complex models, e.g. graph neural networks, and are rarely interpretable. We showed that when using graph features (such as path lengths), you get interpretable predictions and outperform the complex models 🎯 3/4

The paper is the result of my amazing internship at the AutoML Lab in Freiburg, and collaboration with my great coauthors @jovita_lukasik @PetraVidnerova Martin Pilat, Roman Neruda, @mahmoudsafari0 and @FrankRHutter. (credit to Jovita for the cool picture!) 2/4

I’m excited to present you my first paper accepted to a major ML conference - “Surprisingly Strong Performance Prediction with Neural Graph Features” at @ICML 🎉 Blog post: https://t.co/eLvEPxX4Sl Full paper: https://t.co/AurhheEpru 1/4

📢Fully funded PhD position for double-degree at Sorbonne University and Uni of St Andrews, supervised by Nguyen Dang @ndangtt and I topic: using (deep) reinforcement learning for controlling black-box optimization algorithms ℹ️ https://t.co/P0VeinOb9X @StAndrewsCS @LIP6_lab

Separate workspace per research project - a PhD student's dream! Thanks, Vivaldi, for instantly boosting my productivity🔥🙌 #VivaldiBrowser

After three days of attending the #AutoML_Conf in Baltimore @JohnsHopkins, I am left with a thousand new ideas, new friends, and new perspectives on current challenges in the field of AutoML. I can't wait for next year's conference!

I refactored the PyTorch implementation of NAS-Bench-101 training by @HongRomulus into a package installable using pip. If you need to train the nets, but don't want to use TensorFlow, check it out: https://t.co/DvaZNwATjS #AutoML

I am doing a project about elderly programmers. If you are a programmer and over 25 please DM

https://t.co/DlQWZi734S

Pokud se podle někoho svazek dvou lidí stejného pohlaví nesmí jmenovat manželství, protože to je etymologicky špatně, jsem okamžitě pro zákaz horníků, protože ti rozhodně nemůžou z etymologického hlediska pracovat v dole! :)

Nevím, kdo systém [pro rezervaci na očkování proti COVID-19] vyvíjel a testoval, ale jen během první hodiny jsem narazil na následující problémy, píše v diskusi na Lupě @zajdee https://t.co/IORag2zTRA

<3

My se smějem autíčku, že pro maximalizaci objective function "co nejmíň nabourej" stojí na místě, a lidi pak dělaj to samý... :)

A jak je to správně