Clone this wiki locally
Self-Organizing Conference on Machine Learning
The Self-Organizing Conference on Machine Learning is an experimental gathering driven by its participants rather than organizing committee.
Machine learning is moving incredibly quickly. To keep up, many practitioners spend several weeks a year at conferences. However, conference presentations are all on work submitted months prior, meaning that people are already intimately familiar with the content (and it's often already been surpassed).
We'd like to try instead hosting an event focused on the most valuable part of any conference: the people. To do so, we opened a public invitation to machine learning researchers in both academia and industry to attend a new form of conference. The first Self-Organizing Conference on Machine Learning is already oversubscribed but we hope to see a different group of people at a similar event in the future.
SOCML 16 has now concluded. We regard it as a great success.
"I felt like something really different/special was happening at this event, in terms of social dynamics. It felt like an incredibly level playing field: I didn't feel like there was any jockeying for status; everyone at the tables I was at spoke fairly equally; socializing downstairs felt very fluid, not like tight cliques. No-one was showing off, or dismissing anyone" -Catherine Olsson
Participants can remain in contact with each other through the gitter chat room and the discussion forum.
Notes on sessions
The following sessions produced a set of notes recording their discussions:
There were far more sessions than this at SOCML, but most did not produce a written record.
When: Friday and Saturday, October 7-8, 2016, 11am-6pm
**Where: In the OpenAI offices, near the 16th and Mission BART station. The building is wheelchair accessible.
Please bring a photo ID matching your registration to show the check-in desk.
OpenAI will provide lunch and snacks. Participants are responsible for arranging their own travel and lodging.
Completely Unscheduled Rooms
- Forward Prop is a room that is constantly livestreamed to the web. Attendees can enter this room for any conversation they wish to have with web participants. Livestream link here: https://www.youtube.com/watch?v=TKB4zCSjQ7A&feature=youtu.be
- Chain Rule is an area with comfortable couches for a small group discussion.
- Gradient is a microkitchen, appropriate for medium conversations and getting hydrated.
- Backprop, Equilibrium, RMSProp, Variational Autoencoder, and Weight Decay are all small conference rooms that are not scheduled for any predetermined activity. Spontaneously plan a new session in one of these rooms and announce it on the gitter chat and the whiteboard downstairs!
- Kernel Trick, LSTM, Network, Overcomplete, Positive Phase, and Quadrature Pair all desk pods available for spontaneous discussions, just like the conference rooms above.
Attendees are encouraged to bring posters to showcase their recent work. The poster boards will be 4' high x 8' wide.
There will be two poster sessions, one at the start of each day. If you have two posters, we suggest you bring one to each session. If you have only one poster, we suggest you show it at both sessions so that the most people can see it. We anticipate having enough poster boards for everyone to show a poster in both sessions.
Some attendees provided PDFs of their posters to be preserved on this wiki:
- "Improved deep metric learning with multi-class N-pair loss objective" by Kihyuk Sohn and Wendy Shang
- "Perceptual Reward Functions" by Ashley Edwards, Charles Isbell, and Atsuo Takanishi
- "Synthesizing the preferred inputs for neurons in neural networks via deep generator networks" by Anh Nguyen, Alexey Dosovitskiy, Jason Yosinski, Thomas Brox, and Jeff Clune
By attending, you agree to abide by our code of conduct.
We encourage attendees to get to know each other and plan activities for the event using the Gitter chat room.
From a very large pool of applicants, we selected approximately 200 attendees, chosen to ensure a diverse group. We would love to host all of the people who applied, but an unstructured event of this kind works best when limited in size. All applicants were qualified to attend, and we hope to see anyone we were not able to fit this time at a different event in the future.
To provide a somewhat representative sample of attendees, we asked some of the first people who signed up if we could share their names here. Some of these people include:
- Conrado Miranda, PhD student, University of Campinas
- Fei Sha, Associate Professor, UCLA
- Georgia Gkioxari, PhD student at Berkeley
- Leila Wehbe, Postdoctoral Researcher at Berkeley
- Many members of Google Brain, including Samy Bengio, Greg Corrado, and Dumitru Erhan.
- Many members of the OpenAI team.
- Nicolas Papernot, Google PhD Fellow in Security, Pennsylvania State University
- Serena Yeung, PhD student, Stanford University
- Sudnya Diamos, Data Science and Algorithms Engineer, Attune, Inc
- Úlfar Erlingsson, lead of security efforts at Google Research
- Wendy Shang, Software Engineer, Oculus