MIT
Quest for
Intelligence

Robust, Interpretable Deep Learning Systems

November 20, 2018

Quest Symposium on Robust, Interpretable Deep Learning Systems

To advance further, deep learning systems need to become more transparent. They will have to prove they are reliable, can withstand malicious attacks, and can explain the reasoning behind their decisions, especially in safety-critical applications like self-driving cars.

We invite MIT undergraduates, graduate students and postdoctoral scholars to submit current research abstracts to the Quest Symposium on Robust, Interpretable Deep Learning Systems to be held on November 20, 2018. We welcome submissions on attack and defense methods for deep neural networks, visualizations, interpretable modeling, and other methods for revealing deep network behavior, structure, sensitivities, and biases.

The event will feature faculty talks, poster sessions and refreshments.

Speakers

Aleksander Madry
2:30pm. "Robustness and Interpretability" (slides)
Tommi Jaakkola
2:55pm. "Interpretability and functional transparency" (slides)
Stefanie Jegelka
3:25pm. "Robustness in GANs and Black-Box Optimization" (slides)
David Sontag
4:30pm. "Challenges and Dangers of Machine Learning in Health Care"
Luca Daniel
4:55pm. "Evaluating the Robustness of Neural Networks"
Antonio Torralba
5:20pm. "Dissecting neural networks"

Schedule

When: Tuesday November 20, 2018, 2:30pm-6:30pm
Where: Building 46 Atrium and Singleton Auditorium
2:30pm. Aleksander Madry, "Robustness and Interpretability" Singleton Auditorium.
2:55pm. Tommi Jaakkola, "Interpretability and functional transparency" Singleton Auditorium.
3:25pm. Stefanie Jegelka, "Robustness in GANs and Black-Box Optimization" Singleton Auditorium.
3:50pm. Poster Session A. Building 46 Atrium.
4:30pm. David Sontag, "Challenges and Dangers of Machine Learning in Health Care". Singleton Auditorium.
4:55pm. Luca Daniel. "Evaluating the Robustness of Neural Networks" Singleton Auditorium.
5:15pm. Antonio Torralba. "Dissecting neural networks" Singleton Auditorium.
5:40pm. Poster Session B. Building 46 Atrium.
Refreshments will be served.

Call for Posters

Poster format: up to 48''×72'', paper.
Set up posters at 2:00pm before the event.

Symposium Organizers

David Bau, Jun-Yan Zhu, Shibani Santurkar, Aleksander Madry, Antonio Torralba, Fern Keniston
Questions to: ridl@csail.mit.edu
Information at: ridl.csail.mit.edu

The RIDL symposium is affiliated with the Robust Intelligence Initiative @ CSAIL funded by Microsoft and the MIT-IBM AI Watson Lab.