ELLIS Reading Groups
As part of its educational activities, ELLIS aims to connect PhDs and postdocs in the network and supports them in creating opportunities to discuss science.
Currently, two student-led reading groups on machine learning-related topics meet regularly to discuss a paper the group previously voted on. ELLIS Reading Groups are initiated and organized by ELLIS PhD students and postdocs across Europe.
1. Human-Centric Machine Learning (HCML)
The HCML reading group aims to gather researchers and students interested in both getting a wide vision of the topic and also deeply diving into it. Reading papers about different topics inside HCML, and also discussing new problem set-ups, different approaches, and sources of bias will lead us to a broad understanding of how algorithmic and human decisions influence each other.
Chairs: Adrián Arnaiz-Rodríguez, Aditya Gulati, Gergely Németh, Piera Riccio
Next session:
Title: Describing Differences in Image Sets with Natural Language
Authors: Lisa Dunlap, Yuhui Zhang, Xiaohan Wang, Ruiqi Zhong, Trevor Darrell, Jacob Steinhardt, Joseph E. Gonzalez, Serena Yeung-Levy
Abstract: How do two sets of images differ? Discerning set-level differences is crucial for understanding model behaviors and analyzing datasets yet manually sifting through thousands of images is impractical. To aid in this discovery process we explore the task of automatically describing the differences between two sets of images which we term Set Difference Captioning. This task takes in image sets \mathcal D _A and \mathcal D _B and outputs a description that is more often true on \mathcal D _A than \mathcal D _B. We outline a two-stage approach that first proposes candidate difference descriptions from image sets and then re-ranks the candidates by checking how well they can differentiate the two sets. We introduce VisDiff which first captions the images and prompts a language model to propose candidate descriptions then re-ranks these descriptions using CLIP. To evaluate VisDiff we collect VisDiffBench a dataset with 187 paired image sets with ground truth difference descriptions. We apply VisDiff to various domains such as comparing datasets (e.g. ImageNet vs. ImageNetV2) comparing classification models (e.g. zero-shot CLIP vs. supervised ResNet) characterizing differences between generative models (e.g. StableDiffusionV1 and V2) and discovering what makes images memorable. Using VisDiff we are able to find interesting and previously unknown differences in datasets and models demonstrating its utility in revealing nuanced insights.
Presenter: Piera Riccio
When: 24th of September at 3pm CEST
An overview of past and upcoming sessions is provided on the group's webpage.
Join the Google Group here.
2. Mathematics of Deep Learning
The Mathematics of Deep Learning group is for discussing the theoretical background of the state-of-the-art deep neural networks. This theory is aiming at explaining the mechanisms behind the training, generalization, architecture, initialization choosing, etc. This reading group is not restricted by any particular application, so various tasks can be considered, ranging from the purely theoretical linear networks to (current) state-of-the-art transformer networks for natural language processing. The group’s main goal is to keep up with the most advanced areas of research in understanding mathematics behind deep neural networks and their optimization.
Chairs: Linara Adilova (Ruhr University Bochum), Oishi Deb (University of Oxford), Sidak Pal Singh (ETH Zurich)
Next sessions:
Data/Time: Tue, 8th Oct,16:00 CET (15:00 UK time)
Topic: Approaching Deep Learning through the Spectral Dynamics of Weights - https://arxiv.org/abs/2408.11804
Speaker: David Yunis
Affiliation: Toyota Technological Institute at Chicago (TTIC)
Find more information about the group's sessions on this website.
Join the Google Group here.
Who can join ELLIS Reading Groups?
ELLIS Reading Groups are open to any PhD and postdoc students interested in discussing the respective topics with others. New members are always welcome! To join a group, please submit a request via the Google Groups. For any further questions, please contact phd@ellis.eu.
More information
Find more information about the ELLIS PhD Program here. This program has received funding from the European Union’s Horizon 2020 research and innovation programme under ELISE Grant Agreement No. 951847.
Social media
Stay up to date on the latest activities and follow ELLIS on Twitter, LinkedIn, Mastodon and Facebook!
