Science and Engineering of Deep Learning


Science and Engineering of Deep Learning (SEDL) aims to bring about a venue where researchers and practitioners discuss values in machine learning research. We center discussions around two key questions:

  1. What set of scientific and real-world values should we implement to guide the theoretical and practical advances in deep learning?

  2. Why should machine learning researchers be concerned about the broader impact of their research?

Goals of SEDL 2.0

Just like the first edition, “Science meets Engineering in Deep Learning” at NeurIPS 2019, this year, with “Science and Engineering in Deep Learning,” we aim to reach and connect theoreticians and practitioners. In the first iteration of SEDL, we focused on how to bridge communities with seemingly contrasting goals. In the second iteration, we focus on the impact of deep learning in science and engineering applications. This results in two key themes:

  1. Scientific and engineering practices, and

  2. The real-life impact of scientific and engineering practices.

For theme (1), we focus on deep learning as a methodology for scientific discovery. We investigate this theme in the first session, where the cue is ‘values we should implement to guide such advances in deep learning.’ For theme (2), we focus on the social impact of ML research and deep learning as a methodology for scientific discovery. We investigate this theme in the second session, where the cue is ‘broader impact considerations for researchers.’ We finalize the workshop with a panel discussion on values in machine learning.

Workshop Schedule

The workshop will take place on Friday, May 7th 2021 as a virtual event.
Timeline with GMT+1 time (conversion table: SF GMT-7 | NYC GMT-4 | London GMT+1 | Paris GMT+2)

10:30 - 10:45: opening remarks (recorded)
10:45 - 11:45: contributed session

10:45 - 11:05: Samuel J Bell Ideas for machine learning from psychology's reproducibility crisis (recorded talk)
11:05 - 11:25: Jessica Forde and A. Feder Cooper Model Selection's Disparate Impact in Real-World Deep Learning Applications (recorded talk)
11:25 - 11:45: Harshay Shah Do Input Gradients Highlight Discriminative Features? (recorded talk)

12:00 - 14:30: Session 1 - Working towards DL as a methodological tool
12:00 - 12:25: Adyasha Maharana (recorded talk)
12:25 - 12:50: Pushmeet Kohli (recorded talk)
12:50 - 13:15: Joelle Pineau (recorded talk)
13:30 - 14:30: mini-panel & questions (live)
Moderator: Michela Paganini & Session advisor: Nafissa Yakubova

14:30 - 15:00: Break & Poster session

15:00 - 17:30: Session 2 - Social impact of ML research
15:00 - 15:25: Deb Raji (recorded talk)
15:25 - 15:50: Adina Williams (recorded talk)
15:50 - 16:15: Alex Hanna (recorded talk)
16:30 - 17:30: mini-panel & questions (live)
Moderator: Vicente Ordonez-Roman & Session advisor: Emily Denton

17:30 - 18:00: Break & Poster session

18:00 - 19:30: Panel - Values in science and engineering of ML research (live)
Panelists: Danielle Belgrave, Meredith Broussard, Silvia Chiappa, Jonathan Frankle, and Sandra Wachter
Moderator: Shakir Mohamed & Panel advisor: Emily Dinan

Contributed Posters

The following papers were accepted as poster contribution in the workshop:

  1. Dipam Paul (Emory University ), Alankrita Tewari* (KIIT University), Jiwoong Jeong (Emory University), and Imon Banerjee (Emory University) Boosting Classification Accuracy of Fertile Sperm Cell Images leveraging cDCGAN [poster] [paper]

  2. Harshay Shah* (Microsoft Research), Prateek Jain (Google ), and Praneeth Netrapalli (Microsoft Research) Do Input Gradients Highlight Discriminative Features? [paper] Poster session 1

  3. Yu-Lin Tsai* (National Chiao Tung University), Chia-Yi Hsu (National Yang Ming Chiao Tung University), Chia-Mu Yu (National Chiao Tung University), and Pin-Yu Chen (IBM Research) Formalizing Generalization and Robustness of Neural Networks to Weight Perturbations [poster] [paper] Both poster sessions

  4. Arantxa Casanova* (FAIR / Mila), Michal Drozdzal (FAIR), and Adriana Romero-Soriano (FAIR) Generating unseen complex scenes: are we there yet? [video] [poster] [paper] Poster session 1

  5. Hubert Etienne* (Facebook AI) Solving moral dilemmas with AI to address the social implications of the Covid-19 crisis [paper] Poster session 1

  6. Tiffany Cai* (Columbia University), Jonathan Frankle (MIT), David Schwab (Facebook AI Research), and Ari S Morcos (FAIR) Are all negatives created equal in contrastive instance discrimination? [video] [poster] [paper] Poster session 2

  7. Arlene E Siswanto* (MIT), Jonathan Frankle (MIT), and Michael Carbin (MIT) Examining the Role of Normalization in the Lottery Ticket Hypothesis [video] [poster] [paper] Poster session 2

  8. Namhoon Lee* (UNIST), Philip Torr (University of Oxford), and Richard Hartley (Australian National University) Optimal mini-batch size for stochastic gradient methods [poster] [paper] Poster session 1

  9. Camille Ballas* (Dublin City University), César Laurent (Mila, Université de Montréal), Thomas George (MILA, Université de Montréal), Nicolas Ballas (Facebook FAIR), Suzanne Little (Dublin City University, Ireland), and Pascal Vincent (Facebook FAIR & MILA Université de Montréal) Investigating Loss-modelling Pruning Criteria for Unstructured Pruning [video] [poster] [paper] Poster session 2

  10. Samuel J Bell* (University of Cambridge) and Onno P Kampman (University of Cambridge) Ideas for machine learning from psychology's reproducibility crisis [paper]

  11. Arlene E Siswanto* (MIT), Jonathan Frankle (MIT), and Michael Carbin (MIT) Reconciling Sparse and Structured Pruning: A Scientific Study of Block Sparsity [video] [poster] [paper] Poster session 2

  12. Jiaxin Zhang* (Oak Ridge National Laboratory) and Victor Fung (Oak Ridge National Laboratory) Efficient Inverse Learning for Materials Design and Discovery [paper] Poster session 2

  13. Rajiv Movva* (MIT), Jonathan Frankle (MIT), and Michael Carbin (MIT) Studying the Consistency and Composability of Lottery Ticket Pruning Masks Raj Movva [video] [poster] [paper] Poster session 2

  14. Jessica Forde* (Brown University), A. Feder Cooper* (Cornell University), and Michael L. Littman (Brown University) Model Selection's Disparate Impact in Real-World Deep Learning Applications [poster] [paper] Both poster sessions

  15. Saurabh Garg* (CMU), Joshua Zhanson (Carnegie Mellon University), Emilio Parisotto (Carnegie Mellon University), Adarsh Prasad (Carnegie Mellon University), Zico Kolter (Carnegie Mellon University); Sivaraman Balakrishnan (CMU), Zachary Lipton (Carnegie Mellon University), Ruslan Salakhutdinov (Carnegie Mellon University), and Pradeep Ravikumar (Carnegie Mellon University) On Proximal Policy Optimization's Heavy-tailed Gradients [video] [poster] [paper] Poster session 2

The "*" indicates people presenting the work at the poster session. In the list you can also find at which poster session they will participate.

Call for papers

Submissions should be able to express a considerable amount of effort in aligning with the workshop’s themes. In particular, this year, we will be considering submissions in the format of opinion pieces, reviews, or perspectives that are particularly fitting to the theme of the workshop. For technical submissions, beyond scientifically sound arguments and reproducible results, particular attention will also be paid to the accessibility of works by researchers with different backgrounds.

We welcome short but high impact observations and articles covering a wide range of topics, including but not limited to:

  • Investigating practices of deep learning (from dataset creation to modeling) as a tool in scientific discovery

  • Establishing better scientific principles through empirical or theoretical studies of existing algorithms

  • Positive and negative results highlighting the challenges in identifying clear engineering and scientific practices to increase the robustness of our research

  • Proposals for new publication models and analysis of existing peer review systems

  • Bridging the gap between abstract models and social realities

  • Ethical standards guiding the theoretical and empirical advances in the field

  • Defining trade-offs between privacy and transparency of deep learning approaches

  • Overgeneralization, undergeneralization, and the cost of different errors

Submissions should be extended abstracts of no more than 4 pages (excluding references). Abstracts should be appropriately anonymized. All submissions should be in pdf format, must use the ICLR 2021 template (download link for Submissions should be made and managed through the CMT website for SEDL2021. Accepted papers will be presented during poster sessions. About 4 to 6 papers will be selected for contributed talks. Submitted work will be posted on the workshop site under a non-archival status.

Important dates

  • Submission deadline: 5 March 2021

  • Author notification: 26 March 2021

  • Camera-ready papers due: TBD

  • Presentation recordings due: TBD

  • Workshop date: 7 May 2021


  1. Why do you have mini-panels instead of one-on-one Q&As?
    We believe we benefit primarily from bringing people together to discuss a common theme via their own different experiences. As stated in our goal, the workshop focuses on discussing methods, values, and the road ahead, rather than a deep-dive into a specific topic.

  2. What is the advisory role?
    Our workshop relies on prior engagement by all the speakers in each session. Workshop organizers do all the heavy lifting, but we rely on our advisors to help keep the session speakers and moderators in sync and on the topic in the days towards the event.

  3. Is there a difference between SEDL & established workshops focusing on ‘AI for Science’ or ‘Responsible AI’?
    At first sight, the themes we focus on may appear to overlap with existing, established workshops. However, our focus is not on the examples themselves, rather on how we conduct our research and communicate our results.

Speakers and Panelists

Adyasha Maharana
University of North Carolina

Joelle Pineau
Facebook AI Research

Adina Williams
Facebook AI Research

Deb Raji
AI Now Institute

Danielle Belgrave
Microsoft Research

Meredith Broussard
New York University

Sandra Wachter
University of Oxford

Moderators and advisors

Vicente Ordonez-Roman
University of Virginia

Nafissa Yakubova
Facebook AI Research

Emily Dinan
Facebook AI Research


Levent Sagun
Facebook AI Research

Adriana Romero
Facebook AI Research