We are organizing a series of virtual seminars about Automatic Machine Learning (AutoML) as part of the ELLIS units Berlin and Freiburg. The idea is to have bi-weekly, one hour sessions with a 40 minute talk followed by a 20 minute Q/A session. The seminars are free and open to anyone. If you are interested in getting notified the details about the seminars and how to join, please sign up to the group listed below.

Starting from April 2021 we record our sessions and release the videos on our Youtube Channel. Don’t forget to subscribe!

We do not share additional details only to avoid spam, everyone will be accepted in the group and invited to all seminars.

Signup

To get the invites for the seminars, sign up here.

Instructions to join the virtual seminars are available here.

Upcoming seminars

All seminars are happening on Thursday 4-5pm (CET), unless the description specifies a different day/time.

• June 9th – Greg Yang (Microsoft) - Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer
• June 23rd – TBA
• July 7th – Samuel Mueller(University of Freiburg) - Transformers can do Bayesian Inference

Become a speaker

If you want to become a speaker, please send an email to automl-seminars - at - googlegroups - dot - com or contact the organizers.

Organizers

• Giovanni Zappella (Amazon Web Services)
• Aaron Klein (Amazon Web Services)
• Arber Zela (University of Freiburg)
• Jovita Lukasik (University of Mannheim)
• Frank Hutter (University of Freiburg)
• Cédric Archambeau (Amazon Web Services)

Past seminars

• Zi Wang (Google) - Pre-training helps Bayesian optimization too (video)
• Ruohan Wang (UCL) - Global labels and few-shot learning (video)
• Martin Wistuba (AWS) - Few-Shot Bayesian Optimization (video)
• Antonio Carta (University of Pisa) - Avalanche: an End-to-End Library for Continual Learning (video)
• David Salinas (Amazon) - Tutorial on Syne Tune (video)
• David Eriksson (Meta) - “High-Dimensional Bayesian Optimization” (video)
• Jonas Rothfuss (ETH Zurich) -“Meta-Learning Reliable Priors for Bayesian Optimization”
• Willie Neiswanger (Stanford University) -“Going Beyond Global Optima with Bayesian Algorithm Execution”(video)
• Ameet Talwalkar (Carnegie Mellon University & Determined AI) - “Automating Architecture Design for Emerging ML Tasks”(video)
• Xavier Bouthillier (Université de Montréal) - “Reliability of benchmarks and why HPO is important”(ref)(video)
• Bilge Celik (Eindhoven University of Technology), “Adaptation Strategies for Automated Machine Learning on Evolving Data”(video)
• Debadeepta Dey (Microsoft Research) - “Ranking Architectures by their Feature Extraction Capabilities”(video)
• Jörg Franke (University of Freiburg) - “Sample-Efficient Automated Deep Reinforcement Learning”(video)
• Yan Wu (ETH Zurich) - “NAS as Sparse Supernet” & “Trilevel NAS for Efficient Single Image Super-Resolution” (video)
• Louis Tiao (University of Sydney) - “Bayesian Optimization by Classification” (video)
• Gabriel Bender (Google Brain) - “Can weight sharing outperform random architecture search? An investigation with TuNAS” (video)
• Wuyang Chen (University of Texas at Austin) - “Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective” (video)
• Xuanyi Dong (University of Technology Sydney) - Extending the Search from Architecture to Hyperparameter, Hardware, and System (video)
• Josif Grabocka (University of Freiburg) - “Deep Learning for Tabular Datasets”
• Marius Lindauer (University of Hannover) - “Towards Explainable AutoML: xAutoML”
• Arber Zela (University of Freiburg & ELLIS) - “Neural Ensemble Search for Uncertainty Estimation and Dataset Shift”
• Jeff Clune (OpenAI) - “Learning to Continually Learn”
• Binxin (Robin) Ru (University of Oxford) - Interpretable Neural Architecture Search
• Sebastian Flennerhag (Deepmind) - Jointly learning a model and its learning rule
• Colin White (Abacus.ai) - Bananas, Encodings, and Local Search: Insights into Neural Architecture Search (ref1,ref2,ref3)
• Zohar Karnin (Amazon AWS) - Practical and sample efficient zero-shot HPO
• Rodolphe Jenatton (Google Brain) - Hyperparameter Ensembles for Robustness and Uncertainty Quantification
• Matthias Polozeck (Uber AI) - Scalable Bayesian optimization for Industrial Applications
• Joaquin Vanschoren (Eindhoven University of Technology) - Towards ever-learning AutoML
• Frank Hutter (University of Freiburg) - Neural Architecture Search: Coming of Age(slides)
• Valerio Perrone (Amazon): “Fair Bayesian Optimization”(ICML Video)
• Matthias Seeger (Amazon): “Model-based Asynchronous Hyperparameter and Neural Architecture Search”
• Fabio Maria Carlucci (Huawei): Why NAS evaluation is frustratingly hard and how Generator-based optimization can help” (Slides)
• David Salinas (Naver Labs): “A quantile-based approach for hyperparameter transfer learning”