Unofficial Self-Organizing Conference on Machine Learning
When: Saturday, October 1, 2016. 9:00am- 9:00pm
Where: 404 Bryant Street, Sandbox Suites
This event is created for the overflow attendees of the Self-Organizing Conference on Machine Learning by OpenAI.
As you may know, OpenAI is putting together the Self-Organizing Conference on Machine Learning. The number of attendees is limited to about 200 people in order to make the event work best. Members of the community proposed unofficial event for the overflow attendees, so Joseph Catrambone reached us out with this idea asking to help with that.
We are happy to help and would like to thank the community and Joseph for giving us this courtesy. There are 3 parts of unofficial event with the following agenda:
Unofficial SOCML16 is an experimental format of a 12-hours working session consisted of 4 teams tackling Special Projects for OpenAI. Each group is limited to 10 members max. Read the details about the format below.
Meetup for the overflow attendees. This will be a regular gathering for the community, that will be hosted in the evening of unsocml16.
**Schedule:** - 9:00am - registration - 10:00am - welcome note, group formation - 10:15am - 9:00pm - working session - 7:00pm - meetup starts - 9:00pm - closing notes
Groups will be provided with the separate office spaces, in case meetup will be loud.
##Unofficial SOCML16 Format It would be fair to say that this event, as a spin-off of the original conference, is dedicated to OpenAI. It is also fair to say that we, as a community, having diverse backgrounds, expertise and skills can provide meaningful input for them, by tackling or trying to tackle problems they are working on. In that regard, we propose an experimental format, that will allow us to make an impact and showcase our capabilities on both personal and group levels.
The event is focused on Special Projects of OpenAI.
- The Idea is to compose 4 teams accordingly to 4 special projects, that will brainstorm, run experiments, hack or simply put together a report with their findings or conclusions about a particular project.
- The Goal is to provide a new perspective on issues, try to find the solutions or propose a new approach.
- The Outcome of the work of every group will be:
- Code/Repo of proposed solution or results of the experiments that the group has been running
- Report with summary of a session with discoveries, conclusions or assumptions a group has made towards an issue
- Combination of the previous 2 or a 3rd format that will showcase the results of a group
All results will be submitted to OpenAI alongside with the names of the participants and a group lead.
##Prerequisites In order to pull this off, two elements are required:
Group Members are people who compose a team around the topic of interest. Group members are masterminds behind the solutions and findings of the given topics.
Group Lead/Facilitator, is a person who leads the working group and is responsible for the outcome of the team. A group lead possesses necessary skills and expertise in a topic that allows her to guide and navigate through an agenda. To be a facilitator means:
- To understand the goals of the working group
- To keep the group on the agenda and moving forward
- To involve everyone in the meeting, including drawing out the quiet participants and controlling the domineering ones
- To make sure everyone feels comfortable participating
- To develop a structure that allows for everyone's ideas to be heard
- To assign a support person together with whom will capture the working group findings to a written report for immediate delivery.
###Backgrounds This event will assume some familiarity with machine learning, deep learning, and reinforcement learning. Attendees who are not familiar with the concepts below are encouraged to brush up using the references provided below.
- Linear Algebra
- Probability and Programming
- For introductory material on deep Learning and reinforcement learning, see:
- Stanford’s CS231n is a great starting point, covering Convolutional Neural Networks for Visual Recognition
- Books on Deep Learning:
- Deep Learning Book by Ian Goodfellow
- Neural Networks and Deep Learning by Michael Nielsen
- Computer Vision is covered by most, if not all, of the deep learning resources
- Recurrent Neural Networks (RNNs) are the basis of neural network based models that solve tasks related to sequences such as machine translation or speech recognition. Must read: Karpathy’s post and Chris Olah’s post on LSTMs
- Natural Language Processing (NLP): CS224d is an introduction to NLP with deep learning.
- Reinforcement Learning: videos of David Silver on RL Andrej Karpathy’s post on deep reinforcement learning
- Python is a primary choice for DL
##Participation For participation in Unofficial SOCML16 fill in the registration form below: UNSOCML16 Registration Please choose how do you want to participate (Group Lead/Group Member) and the topic you are interested in. ####Choosing a Group Lead A group lead is required to have the relevant background, expertise, and skills and is responsible for a group outcome. We use our network and reaching out people who can be a good fit for that, but we would like to encourage members of the community to apply as well as propose someone who might be a good fit (send your proposals to email@example.com)
All 4 group leads will be announced prior the event in the Slack channel
For participation in the meetup, RSVP here