"Causal Deep Learning"
Jeroen Berrevoets, Krzysztof Kacprzyk, Zhaozhi Qian, Mihaela van der Schaar

Prof Mihaela van der Schaar just gave a very exciting talk at the #counterfactuals workshop at #icml2023 on her lab's CDL framework dealing explicitely and with concepts from #causality and #machine-learning. Definitely going on my reading list.

https://arxiv.org/abs/2303.02186

Today, Michael Hu is presenting our work at the High Dimensional Learning Dynamics workshop at #ICML2023. It has the greatest visualizations I’ve ever touched. It’s the Mona Lisa of training dynamics posters. There’s gonna be a corner of the gallery full of weeping empiricists mumbling “only art is real”. Go see these plots and tell your grandkids. https://michahu.github.io/assets/pdf/latent_state_dynamics.pdf

Excited to present my poster about regression on Grassmann manifolds for the analysis of multi-condition single-cell data at #ICML2023 in the TAG workshop at 11:10am :)

Read the extended abstract at openreview.net/pdf?id=MrE4jL0… and check out the software at https://github.com/const-ae/lemur

Check out this amazing Google DeepMind paper presented at [ICML] Int'l Conference on Machine Learning #ICML2023!

"Can Neural Network Memorization be localized?"

Memorization in neural networks is a process where a specific set of neurons in different layers of the model store and remember specific patterns or data.

Hanie Sedghi et al conducted groundbreaking research using a technique called "Example-tied Dropout." This approach allows them to pinpoint and limit memorization to certain fixed neurons that can be discarded during the testing phase.

Read more here>https://arxiv.org/pdf/2307.09542.pdf

#icml2023
#posterguerilla

#ICML2023 test of time award (most impactful papers from in #MachineLearning from ICML 2013):
https://icml.cc/Conferences/2023/Test-of-Time

- Learning Fair Representations - which started a whole new subfield (shameless plug that Sevvandi has a #Job #Opening for a #postdoc in #Australia https://jobs.csiro.au/job/Melbourne%2C-VIC-CSIRO-Postdoctoral-Fellowship-in-Fairness-Research-in-Machine-Learning-%28FairML%29/941838510/)

The two runner ups:
- multimodal representation learning
- AutoML, hyperparameter search.

#ICML2023 outstanding papers award (6 of them):
https://icml.cc/Conferences/2023/Awards

- theoretical analysis of non-smooth stochastic convex optimization
- watermarking output of large language models
- learning Boolean functions, generalising to unseen ones
- Follow the Regularized Leader (FTRL) variations for optimising imperfect information games
- Markov Chain Monte Carlo for graphs - via self-repellent random walks
- Adaptive experimental design - a new objective called algorithmic information ratio

About 65% of #ICML2023 attendees are wearing shorts or something similar 🩳🏖️

The 40th edition of the international conference for researchers on Machine Learning is currently holding at the Hawaii Convention Center, USA from July 23rd to 29th.

Here are the six (6) papers that received Outstanding Awards at #ICML2023:

1. Learning-Rate-Free Learning by D-Adaptation
Authors: Aaron Defazio et al.
Link: https://openreview.net/forum?id=GXZ6cT5cvY

2. A Watermark for Large Language Models
Authors: John Kirchenbauer et al.
Link: https://openreview.net/forum?id=aX8ig9X2a7

3. Generalization on the Unseen, Logic Reasoning and Degree Curriculum
Author: Emmanuel Abbe
Link: https://openreview.net/forum?id=3dqwXb1te4

4. Adapting to game trees in zero-sum imperfect information games
Authors: Come Fiegel et al.
Link: https://openreview.net/forum?id=O1j4uFuSVW

5. Self-Repellent Random Walks on General Graphs - Achieving Minimal Sampling Variance via Nonlinear Markov Chains
Authors: Vishwaraj Doshi et al.
Link: https://openreview.net/forum?id=450iImFM4U

6. Bayesian Design Principles for Frequentist Sequential Learning
Author: Yunbei Xu
Link: https://openreview.net/forum?id=tRhQsHnoFw

Lectures, poster sessions & plenty of productive discussions: Some impressions from the #ELLISUnConference currently taking place at HEC Paris! Participants can also connect live to #ICML2023 tutorials.

#HiPARISCenter #AI #ML

More ➡️ https://ellisunconference2023.github.io

Slides for the #icml2023 tutorial on "Trustworthy Generative AI" by Nazneen Rajani, Hima Lakkaraju, and Krishnaram Kenthapadi
Tutorial website: https://sites.google.com/view/responsible-gen-ai-tutorial
Slides: https://t.co/hNQMkXOqgZ
@nazneenrajani @ICMLConf #trust #trustworthyAI #AI #icml #icml2023
#icml23 #generativeAI

@shakir using GANs to predict heavy rainfall. Requires more suitable eval metrics for ML, backed up by human expert evaluation
#icml2023

@MarzyehGhassemi speaking about ‘Taking the Pulse of Ethical ML in Health’ as the first keynote of #ICML2023 🤩 Have diverse data, audit models, deploy fair advice. So many insights in the progress on ML and Health. 👏👏

#icml2023 Marzyeh Ghassemi keynote: using demographic features can sometimes result in worse predictions for some groups of people. Well, yeah. EBMs can help expose and make it easy to adjust for this

#icml2023 Marzyeh Ghassemi keynote: phrasing of the label makes a big difference. Eg. description vs does it surpass a threshold

#icml2023 Marzyeh Ghassemi keynote: AI can predict race better than humans, from the data. To improve better outcomes for all

#icml2023 best papers. Congrats!
https://icml.cc/Conferences/2023/Awards

#ICML2023 attendance overview

#icml2023 attendance by geography

#ICML2023 #whova app doesn't let you add masotodon ad a social?