Embedded Files
Automatic Machine Learning, Deep Learning, and GPUs
Stanford Alumni Center, August 11, 2015
In conjunction with the INNS BIG DATA conference in San Francisco, August 8-10, 2015.
See our souvenir album (thanks to Sebastien Treger)
Realize your dream: building the perfect machine learning black box, which can learn without any human intervention!
The goal of this event is to form teams in the Bay Area to join the AutoML challenge, with $30,000 in prizes donated by Microsoft. The participants will have the opportunity to familiarize themselves with Codalab, a new collaborative platform, deep learning, and GPUs.
INSTRUCTIONS
==> We have updated the [GPU TRACK INSTRUCTIONS]. You may want to get familiar with them before the event.
==> The [BOOTCAMP SUBMISSIONS] website is now open for submissions. You have until midnight on August 11 to make your submissions. After that, we urge you to try to enter the AutoML challenge.
ACCEPTED REGISTERED PARTICIPANTS
The participants are organized in groups. The group name is the Codalab ID. It is possible to communicate with the members of the group by writing to <group-name>@chalearn.org.
PROGRAM
8:30 am: Registration
8:45 - 9:30 am: Isabelle Guyon - Presentation of the AutoML challenge. [slides]
9:30 - 10:15 am: Arthur Pesah - Introduction to Python machine learning libraries. [iPython notebook -- start with > ipython notebook Arthur_notebook.ipynb]
10:15 - 10:30 am: Break.
10:30 - 11:15 am: Julie Bernauer. NVIDIA - Introduction to GPUs. [slides]
11:15 am - 12 am: Percy Liang - CodaLab Worksheets for Efficient Collaborative Research. [slides]
12:00 am - 1 pm - "Free" lunch.
1:00 pm - 1:30 pm: Sebastien Treger - Walk through the submission process and the sample code.
1:30 pm - 4:30 pm - Hackathon/Advanced bootcamp: work on making submissions to the AutoML challenge with the help of coaches. [BOOTCAMP SUBMISSIONS][GPU TRACK INSTRUCTIONS]
4:30 pm - 5:00 pm - Top ranking participants share their solutions.
ABSTRACTS
Presentation of the AutoML challenge (Isabelle Guyon)
The AutoML challenge offers an unprecedented opportunity for the community of machine learning to try to create a perfect black box capable of training learning machine for classification or regression from any feature-based dataset, without any human intervention. The presentation will go over the challenge protocol and give guideline on how to make successful entries by going over strategies for hyper-parameter selection and overfitting avoidance and reviewing some of the strategies of winners of previous rounds.
CodaLab Worksheets for Efficient Collaborative Research (Percy Liang)
We are interested in solving two infrastructural problems in data-centric fields such as machine learning: First, an inordinate amount of time is spent on preprocessing datasets, getting other people's code to run, writing evaluation/visualization scripts, with much of this effort duplicated across different research groups. Second, a only static set of final results are ever published, leaving it up to the reader to guess how the various methods would fare in unreported scenarios. I will present CodaLab, a new platform which aims to tackle these two problems by creating an online community around sharing and executing immutable components called bundles, thereby streamlining the research process. Finally, I will show how worksheets can be useful for participating in competitions.
LINKS
- Codalab
- Prerequisites
- Create your own Amazon GPU server
- Introduction paper (accepted to IJCNN 2015)
- GPU track instructions
- NVIDIA machine learning
- NVIDIA deep learning course
- Sklearn Kaggle examples
ORGANIZERS
Isabelle Guyon, ChaLearn (contact: events@chalearn.org)
Pauline Essalou, NVIDIA
Emilia Vaajoensuu, NVIDIA
Percy Liang, Stanford University
Arthur Pesah, ChaLearn
Sebastien Treger, ChaLearn
Lukasz Romaszko, ChaLearn
Evelyne Viegas, Microsoft
COACHES
Arthur Pesah, ChaLearn
Sebastien Treger, ChaLearn
Eric Carmichael, Tivix
Francis Cleary, Tivix
Julie Bernauer, NVIDIA
LOCATION
Lane/Lyons/Lodato room in Fisher Hall
Frances C. Arrillaga Alumni Center
Arrillaga Alumni Center
326 Galvez St
Stanford, CA 94305
United States
PARKING
GALVEZ LOT, CODE = 8927
We recommend that you park in the Galvez Lot. You can get a discount parking for this event.
1. First park and remember your stall number
2. Touch any key to exit intro screen
3. Enter your stall number and Press [OK] to Continue
4. Select #1 to Purchase a Ticket.
5. On the next screen, select (1) for Incremental or (2) for All Day Parking. If arriving after 10:45 use the incremental option without an event code and skip to step 9.
6. If All Day Parking is selected, select "Yes" when it asks if you have an event code
7. Enter the event code number (8927) and press OK
8. Once the code is entered, you'll receive a message that says "Get Ready For Payment"
9. Insert your Visa/Mastercard or cash payment (if incremental rate was selected and paying by credit card, increase the expiration time by pressing the '1' key). If paying by cash, depositing more money increases the expiration time
10. Select OK to complete transaction and print your receipt. No need to display receipt on dashboard
See your other parking options. At the lower right corner of the map below you see "Parking & Transportation": this is where you can get your one-day pass: 340 Bonair Siding.
See full souvenir album (thanks to Sebastien Treger)
Page updated
Report abuse