Machine Learning for Discovery in Particle Physics

What are Dark Matter and Dark Energy?  Why is our universe only made of matter when there were equal amounts of matter and antimatter involved in the big bang?  Why are some fundamental particles, like the top quark which weighs almost as much as a gold nucleus, so heavy while others, like neutrinos so light that we haven't even managed to measure their mass yet?  These are some of the big, open questions facing particle physics.  At the CERN lab, in Geneva, Switzerland, the Large Hadron Collider (LHC) collides protons together at the highest energies ever achieved in a laboratory, searching for signs of new particle or new interactions that would help us answer those questions.  Notre Dame participates in the Compact Muon Solenoid (CMS) experiment, that collects data from these LHC collisions.  This research project involves analyzing CMS data to search for signs of new particles or interactions.

While the answers we seek require new particles and/or interactions, there is no guarantee that those particles exist within the energy reach of the LHC.  The LHC converts the kinetic energy of accelerated protons into new particles, using Einstein's famous E=mc^2 equation.  However, particles whose mass exceeds the available proton kinetic energy in the collisions cannot be produced directly.  Because it will be a long time before we have a more powerful particle accelerator, my group is using a technique called Effective Field Theory (EFT) to search for indirect signs of very heavy particles might have on the LHC data.  EFT allows us to predict how different types of hypothetical super-massive particles might impact the CMS data, and then search for those effects.  

Because the indirect effects we're looking for may be small, and because the data is complex and rich, we need to turn to advanced analysis techniques to maximize our chances for discovery.  The most powerful technique available is known as machine learning, when our computational algorithm learns for examples of simulated data how to spot the interesting collisions from among all the uninteresting ones.  The specific tool we use is called a deep neural network.

In this project you will analyze CMS data, looking for ways to improve sensitivity to EFT effects through the application of deep neural networks.  There may also be the opportunity to work on machine learning techniques to enhance the statistical analysis or visualization of high dimensional data.

Name of research group, project, or lab
The Lannon Group
Why join this research group or lab?

In joining the Lannon Group, you will have the opportunity to become part of a team of postdoctoral researchers, graduate students, and other undergraduate students who apply advanced data analysis techniques to try to tease out nature's secrets from some of the most complex data ever collected in a scientific experiment.  You will have the opportunity to interact with the international CMS collaboration, networking with researchers working on the CMS experiment around the globe.  You will have the chance to work with cutting edge data science and machine learning tools.  Your efforts will help to advance humanities understanding of the fundamental laws of nature that govern our universe.  It is highly unlikely, at any point during this research, that you would fall through the event horizon of a black hole and be lost from the space-time continuum forever.

Representative publication
Logistics Information:
Project categories
Physics & Astronomy
Student ranks applicable
First Year
Sophomore
Junior
Senior
Student qualifications

Required: Ability to write code in Python and basic computing proficiency, including how to access the terminal.

Desirable:  The following are not required, but would be useful to know.  If you are selected for the project and don't know any of these skills you will learn them.

  • Familiarity with Numpy and/or ROOT
  • Knowledge of special relativity concepts like Lorentz transformations or invariant mass.  (Taught in Physics C.)
  • Knowledge of basis probability and statistics concepts.
  • Some knowledge of linear algebra
  • Familiarity with distributed computing systems, for example, using ssh to log into a remote computer.
  • Experience with deep neural networks, especially using Pytorch
  • Some familiarity with the basic concepts of quantum mechanics, as taught in Physics D.
Hours per week
1 credit / 3-6 hours
2 credits / 6-12 hours
Compensation
Research for Credit
Unpaid - Volunteer
Number of openings
2
Techniques learned

By engaging in this project, you will learn the following skills and techniques:

  • Advanced python programming
  • Distributed computing including the use of batch systems
  • Data science skills
  • Basic experimental particle physics skills, including how to analyze particle physics data
  • Basic familiarity with Feynman Diagrams
  • Probability and Statistics
  • Making presentations and communicating with a diverse group of people
This project will use an Expectations and Structure agreement.
Expectations and Structure

In agreeing to join this project, you are agreeing to commit to the following:

  • You will commit to a one-hour weekly meeting.  The meeting may be one-on-one or a group meeting depending on the number of people collaborating on the project and the project phase.  The weekly meeting will be canceled any time the day of the meeting falls on a class holiday (e.g. Thanksgiving break, Spring Break, etc.)
  • You will commit to working on the project independently outside of the group meeting for an amount of time appropriate to the time commitment agreed upon at the start of the project.  For projects undertaken for research credit, this would be 3-6 hours outside of the weekly meeting per research credit hour.  For research undertaken on a voluntary basis, the amount of time will be agreed upon between you and me at the start of research.
  • You will communicate about any changes in your schedule in advance, in particular, if you have a need to miss a weekly meeting.
  • You will not commit scientific fraud or misconduct.  This means you will not falsify data or claim someone else's work as your own.
  • You will maintain a professional demeanor towards all other members of the Lannon Group or any other student, faculty member, staff member, or researcher that you interact with on Lannon Group business.  This means that you will be polite and patient in your interactions and will never take an action intended to intimidate, demean, insult, or harm another person.
  • You will speak up if at any point you feel like you have experienced a negative interaction with another member of the Lannon group, including myself.  If you do not feel comfortable talking to me, you will approach another member of the Physics Department Faculty or staff, such as the Undergraduate Program Coordinator, the Associate Director for Undergraduate Studies, or the Director for Undergraduate Studies.
  • To present your research at least once per year at either FURF or COS-JAM.

When you join this project, I agree to the following:

  • To provide you with guidance and help in pursuing your research, either directly, or by putting you in contact with an appropriate additional mentor, such as a postdoctoral researcher, graduate student, or another undergraduate student.
  • To communicate with you whenever my travel schedule or other university or personal commitments will necessitate missing a meeting.
  • To provide you with the necessary resources to pursue your research project, within the limits of available funding.
  • To provide the opportunity to present any completed and approved research results at a national or international conference.
  • To treat you with the same professional manner, including patience, respect, and kindness that I expect you to show to everyone else in the group.
  • To respond promptly to any concerns raised about the research group environment, regardless of the source of those concerns.

If you pursue research for credit, to earn a grade of an "A" you will be expected to do the following:

  • Meet at the start of the semester to set one or more concrete research objectives for the semester, depending on the number of credits.
  • Present an update of your progress each week during our regularly scheduled meeting.
  • Meet with me if you become concerned that you will not be able to achieve any of the objectives originally planned and work with me to set new expectations, if appropriate.
  • Provide a brief written summary of your accomplishments before the end of the final exam period.

If you are pursuing research for credit free to leave my group at the completion of any semester.  If you want to leave during the semester, you will need to abide by university rules for dropping or withdrawing from a class.  Ordinarily you will be invited to continue to research in my group from one semester to the next.  However, I reserve the right to decline to continue research with you if our collaboration is not working out.  I will let you know at the end of the semester if this is the case.  If you have not informed me of your intention to continue research into the next semester by the last day of the current semester, I reserve the right to assume that you intend to seek opportunities elsewhere.

If you are pursuing research on a voluntary basis (i.e. no course credit), you can stop doing research at any time.  I would, however, appreciate that you meet with me to communicate your intention to leave my group in advance.

Contact Information:
Mentor
klannon@nd.edu
Professor
Name of project director or principal investigator
Kevin Lannon
Email address of project director or principal investigator
klannon@nd.edu
2 sp. | 1 appl.
Hours per week
1 credit / 3-6 hours (+1)
1 credit / 3-6 hours2 credits / 6-12 hours
Project categories
Physics & Astronomy