Involving other communities through challenges and cooperation

http://lhcathome.web.cern.ch/projects/atlas

ATLAS@Home

Content from Claire Adams Bourdarios

Various badges for users to earn through their contributions.

• The ATLAS@home project allows volunteers to run simulations of collisions in the ATLAS detector using the idle processing time of their home computers.
• For the first year the community consisted of software fans, who were attracted by the technical challenge and contributed to the debugging via message boards.
• With the start of LHC, the number of people attracted for outreach reasons is growing.

World map of users by credit contributed:

Higgs Hunters

www.higgshunters.org

Content from Alan Barr

Zooniverse is a collection of web-based citizen science projects that use the efforts of volunteers to help researchers deal with the flood of data that confronts them.

Rewards are for different contributions:

• Quarks - for being in the top Nth percentage of the users contributing the most Recent Average Credit (RAC).
• The top 1% get a top quark.
• Leptons - for perseverance.
• After 180 days, they get a tau neutrino.
• Bosons - years of involvement.
• After 5 years, they get a Higgs.

ATLAS jobs running on BOINC site:

ATLAS Simulation Jobs:

No knowledge of particle physics is required, but for those interested in the physics processes simulated in ATLAS@Home as well as the ATLAS experiment itself visit our public page.

Greater than one million users worldwide!

With 20 projects and 60 papers.

"A new project! Yes!" - zombie67 (First user), 18th June 2014

"More WUs please !" - Yeti, 31st January 2015

Interesting events seen so far...

WU = Work Unit

Higgs Hunters went live on Zooniverse in November 2014

C. Nellist

4 / 8

EPS - HEP 2015 (22-29th July 2015), Vienna, Austria

Q: What are they classifying?

A: "Lines that seem to sprout from a common point that is NOT the center. These are called "Off-center vertexes" (OCVs)."

The statistics:

A 'punch-through'

A high-energy particle that bursts into a jet of particles at the edge of the calorimeter and looks like a jet of muons.

Possible 'beam halo'

More than 800,000 classifications of 85,000 images by more than 6,000 different members of the public.

Particles that have been swept along with the LHC beam.

5 / 8

C. Nellist

EPS - HEP 2015 (22-29th July 2015), Vienna, Austria

Introduction

The ATLAS collaboration has recently setup three projects targeting citizen science or specific communities.

• Higgs Machine Learning
• ATLAS@Home
• Higgs Hunters

Contributors

ATLAS@Home

Higgs Machine Learning

Andrej Filipčič

Claire Adam Bourdarios

David Cameron

Efrat Tal Hod

Eric Lancon

Riccardo Maria Bianchi

Wenjing Wu

ATLAS physicists :

LAL: Claire Adam-Bourdarios

David Rousseau

RHUL: Glen Cowan

Machine Learning scientists :

LAL: Balazs Kegl

LRI: Cécile Germain

Chalearn: Isabelle Guyon

C. Nellist

EPS - HEP 2015 (22-29th July 2015), Vienna, Austria

2 / 8

Higgs Hunters

Further reading:

http://www.symmetrymagazine.org/article/july-2014/the-machine-learning-community-takes-on-the-higgs

http://www.symmetrymagazine.org/article/november-2014/needed-citizen-scientists-for-higgs-hunt

Oxford: Alan Barr

Thomas Hornigold

Chris Lintott

Ryan MacDonald

NYU: Andy Haas

Jeffrey Mei,

Birmingham: Pete Watkins

C. Nellist

EPS - HEP 2015 (22-29th July 2015), Vienna, Austria

8 / 8

Communicating the projects

Social media can be used to inform the public about our various projects and encourage participation and discussion.

#HiggML

Higgs Machine Learning

http://higgsml.lal.in2p3.fr

Artwork by Sandbox Studio, Chicago. From Symmetry Magazine.

Posted on Twitter, 9:29 AM - 17th June 2014

Content from David Rousseau

Goal: Sort into two groups.

1) Higgs

2) Everything else

Aim: To benefit from the knowledge of the machine learning community and to engage the public with the type of analysis work required to discover the Higgs.

Duration: The challenge ran from May to September 2014 on the Kaggle platform.

Sample: The signal used was Higgs to two tau particles, simulated by the ATLAS experiment.

Posted on Twitter, 9:29 AM - 27th November 2014

Prizes:

1st Place - $7,000

2nd Place - $4,000

3rd Place - $2,000

CERN Ask Me Anything (AMA) on Reddit, June 2015

The HiggsML poster advertising the challenge.

Outcome:

• The significance of the dataset increased by 20% with respect to the traditional HEP tool (TMVA).
• It was a very fruitful collaboration established with the machine learning community.
• New techniques and new software came from the collaboration.

EPS - HEP 2015 (22-29th July 2015), Vienna, Austria

C. Nellist

6 / 8

Tianqi Chen and Tong He received the "HEP meets ML" award

The dataset used in the challenge will remain on CERN Open Data Portal with a citeable d.o.i.:

http://opendata.cern.ch/collection/ATLAS-Higgs-Challenge-2014

• Contains ~800k events.

Kaggle’s most popular challenge event at the time!

3 / 8

C. Nellist

EPS - HEP 2015 (22-29th July 2015), Vienna, Austria

How you can get involved?

Although the HiggsML challange on Kaggle has finished, the dataset is still available online: http://higgsml.lal.in2p3.fr

• LHCb have just released their own challenge to ask people to identify a rare decay phenomenon:

https://www.kaggle.com/c/flavours-of-physics

• ATLAS and CMS are looking at another challenge.

Instructions for downloading ATLAS@Home can be found on the website:

http://atlasathome.cern.ch

For Higgs Hunters, you can go directly to this link and get started!

http://www.higgshunters.org/

EPS - HEP 2015 (22-29th July 2015), Vienna, Austria

7 / 8

C. Nellist

Involving other communities through challenges and cooperation

Clara Nellist (LAL-Orsay)

on behalf of the ATLAS Collaboration

EPS - HEP 2015 (22-29th July 2015), Vienna, Austria