So there is this new (short) video by Axel Plinge on our paper Recipes for Post-training Quantization of Deep Neural Networks at youtu.be/wyEJzUyuGGA

Lilian Wang Lilian Weng asked "Are Deep Neural Networks Dramatically Overfitted?" and touches many fundamentals, from Occam's Razor over Universal Approx. Theorem to Intrinsic Dimensions and the Lottery Ticket Hypothesis. Lil'Log is always worth reading! lilianweng.github.io/lil-log/2019/0…

TIL: In a distributed ML setting, noise in a communication channel (i.e., between split network layers that run inferences) could be simply interpreted as Gaussian dropout! Thanks for the inspiration Simone Scardapane

A short summary / interesting read of Kaparthy’s talking points from CVPR’21 on “why self-driving cars don’t need lidar” (via VentureBeat) venturebeat.com/2021/07/03/tes…

Our paper on accelerating similarity-based queries for soccer was just published in the ACM Transactions on Intelligent Systems and Technology! Is metric learning your thing? Got trajectory data? Like soccer? This is for you! dl.acm.org/doi/10.1145/34…

I am happy to visit Christopher Rozell's SIPLab at Georgia Tech for an exciting research collaboration in the coming months. This environment is highly conducive to ideas! #FraunhoferIIS #ADA #FAU

"I am writing regarding your JMLR manuscript entitled "...." I am pleased to inform you that your manuscript has been accepted for publication." Feels really good 😎 Christoffer Löffler Journal of Machine Learning Research

I can totally recommend reading Luca's article on Searching for Soccer Scenes using Siamese Neural Networks (in Towards Data Science) towardsdatascience.com/searching-for-…

We've recently benchmarked eight saliency methods for conv. networks on time-series classification and segmentation. Turns out, there is no winner for all combinations of {model, metric, dataset}. Preprint: arxiv.org/abs/2203.07861 Code: github.com/crispchris/sal… #xai #saliency

Our paper "IALE: Imitating Active Learner Ensembles" is now available on JMLR! 😀 Check it out! Christoffer Löffler jmlr.org/papers/v23/21-…

Check out our new preprint on active learning for ordinal embeddings: arxiv.org/abs/2207.12710 We've collaborated with our friends at SIPLab (Georgia Tech) to learn similarity from few annotations.

#pelicansgameday with Christopher Mutschler

Visit us today at 11am at poster #1008 at #NeurIPS2022. We’ll present our work on Deep Active Learning. For convenience, just start your poster session hiking trip at the end of the hall. Christopher Mutschler neurips.cc/virtual/2022/p…

Probably the winner for most creative poster at #NeurIPS2022 used 8 iPads to bring the point across

Transformers are eating continual learning (datasets)! #NeurIPS2022 workshop by Paul Janson, Wenxuan Zhang, Rahaf Aljundi and Mohamed Elhoseiny

I am very happy that our paper finally made it into TMLR! Christoffer Löffler

Our user study on Active Learning of ordinal embeddings got published in Transactions on Machine Learning Research! See our video at youtu.be/xqOJAtjxjKE for the full story of how we learn similarity functions for football tracking data. Christopher Mutschler Bjoern Eskofier Christopher Rozell Dario Zanca

In our work on Active Learning of Ordinal Embeddings¹ we joined forces with GaTech's SIPLab to improve the search on trajectory data² with the InfoTuple Active Learning method³ ¹ openreview.net/forum?id=oq3tx… ² dl.acm.org/doi/10.1145/34… ³ doi.org/10.1609/aaai.v