iiinova — where your digital self keeps evolving

Aurelien Lucchi @alucchi.bsky.social · 03/05/2025

Post-doc position in the field of Optimization and Deep Learning Theory jobs.unibas.ch/offene-stell... PhD position in the field of Optimization and Deep Learning Theory jobs.unibas.ch/offene-stell... PhD position in the field of Reasoning in Machine Learning jobs.unibas.ch/offene-stell...

jobs.unibas.ch

Universität Basel: Post-doc position in the field of Optimization and Deep Learning Theory

The Optimization of Machine Learning Systems Group (Prof. A. Lucchi) at the Department of Mathematics and Computer Science at the University of Basel is looking for one post-doctorate to work in the a...

061

Aurelien Lucchi @alucchi.bsky.social · 03/05/2025

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.

jobs.unibas.ch

Universität Basel: Post-doc position in the field of Optimization and Deep Learning Theory

1106

Reposted by Aurelien Lucchi

Clément Canonne @ccanonne.github.io · 22/04/2025

Insightful pair of blog posts by @stein.ke on #privacy amplification by subsampling: instead of running your privacy-preserving algo on the full dataset, run it on a random subset. Better!(?) Why, and what the limitations? differentialprivacy.org/subsampling/ differentialprivacy.org/subsampling-...

differentialprivacy.org

Privacy Amplification by Subsampling

Privacy Amplification by Subsampling is an important property of differential privacy. It is key to making many algorithms efficient – particularly in machine learning applications. Thus a lot of wor...

1142

Aurelien Lucchi @alucchi.bsky.social · 11/12/2024

Joint work with Rustem Islamov · Niccolò Ajroldi · Antonio Orvieto

020

Aurelien Lucchi @alucchi.bsky.social · 11/12/2024

Today @NeurIPSConf 2024 at 4.30pm: Loss Landscape Characterization of Neural Networks without Over-Parametrization. West Ballroom A-D #6008 We propose a novel class of functions that characterize the loss landscape of deep models without requiring over-parametrization. arxiv.org/pdf/2410.12455

150

Aurelien Lucchi @alucchi.bsky.social · 09/12/2024

Only a 70B version, nothing smaller?

110

Reposted by Aurelien Lucchi

Helen Czerski @helenczerski.bsky.social · 08/12/2024

Ever wondered why walking around museums is weirdly tiring? At normal speed our legs act like pendulums, swinging forward from the hip & saving us a huge amount of energy. In "museum shuffle" our muscles must do ALL the work of constant readjustment. So cake in the cafe is scientifically justified 🥳

744251662800

Aurelien Lucchi @alucchi.bsky.social · 26/11/2024

Also would like to be in this list :)

100

Aurelien Lucchi @alucchi.bsky.social · 23/11/2024

Lots of snow in Switzerland this week-end! Even our usually reliable train system was caught by surprise.

040

Aurelien Lucchi @alucchi.bsky.social · 19/11/2024

Swiss French here and I do ML. By definition we are of the passive agressive/grumpy kind.

000