STRUCTURED. EVIDENCE-BASED. BUILT FOR PHYSICISTS.

Machine Learning for High Energy Physics

A self-paced course that takes you from your first question about machine learning to a result your collaboration will trust — explained in plain language, built on published physics research.

Get Module 1 + the Roadmap — Free →

Free Module 1 and the one-page Roadmap. No payment, unsubscribe any time.

Built on peer-reviewed physics researchReal collision data and simulation30-day money-back guarantee

Start here

Which of these sounds like you?

Pick the one closest to where you are today. Each path leads to the same free starting point, tuned to what you need most.

New to Machine Learning

New to Machine Learning

Your supervisor asked you to use machine learning in your analysis, and you have no idea where to start.

Start with Module 1 →
 Getting More From Your Cuts

Getting More From Your Cuts

You have pushed cut-based selection as far as it will go, and you need machine learning to find the signal your cuts keep missing.

Start with Module 1 →
 Reviewing an ML Analysis

Reviewing an ML Analysis

You have to judge whether a colleague's machine-learning result is solid enough to put your name behind and publish.

Start with Module 1 →
Working through a physics analysis on a laptop

How it works

Three steps to a result you trust

1

Start free

Get Module 1 plus the one-page Roadmap — the exact path from where you are today to a finished analysis, sent straight to your inbox. No cost, no catch.

2

Follow the six modules

Each module is short, plain-spoken, and hands-on. You work with real collision data and simulation, one step at a time, in the order you would actually build an analysis.

3

Get a result you can defend

Finish with an analysis you understand end to end — and the answers to every question a review committee will ask before it lets you publish.

The evidence

Why bother learning this at all?

Because the published record is clear: done properly, machine learning finds more of your signal and stands up in review. Here is some of what it shows.

On the kind of separation problems physicists face every day, machine learning has been shown to keep far more of the signal you care about while throwing away more of the background — a large, measurable gain over hand-tuned cuts.

Roe et al., Nuclear Instruments and Methods A 543 (2005) 577–584

The same methods have topped open physics challenges, including the Higgs Machine Learning Challenge, where thousands of teams competed to pull a real signal out of simulated LHC data.

Adam-Bourdarios et al., JMLR W&CP 42 (2015); Chen & Guestrin, KDD 2016

Toolkits like TMVA made these methods a standard part of experimental physics, and the workflow they introduced is still the reference many analyses follow today.

Hoecker et al., arXiv:physics/0703039 (2007)

Deep neural networks trained on raw detector data can match or exceed the sensitivity of hand-engineered features for distinguishing new-physics signatures from Standard Model backgrounds.

Baldi, Sadowski, & Whiteson, Nature Communications 5, 4308 (2014)

Machine learning is now used across the full LHC workflow — from real-time triggering and event reconstruction to final statistical inference — and has become essential to extracting physics from the data volumes produced by Run 2 and beyond.

Radovic et al., Nature 560, 41–48 (2018); Guest, Cranmer, & Whiteson, Annu. Rev. Nucl. Part. Sci. 68 (2018)

A finished analysis plot ready for review

The payoff

The analysis you meant to build — finished, and defensible

Not a black box you cannot explain to your convener. A result you understand from end to end: you know why it works, what it is sensitive to, and how to answer every question the review committee throws at it.

Every module builds on the last, in the same order you would actually do the work — from your first look at the data to the final plots that get an analysis approved.

Get Module 1 free →

Free · 30-day money-back guarantee on the full course

What "evidence-based" actually means here

Most machine-learning tutorials are written for software engineers, with examples about spam filters and photos of cats. This is not that. Every module is built for physicists, uses the kind of data you already work with, and is grounded in the peer-reviewed physics literature — the same papers your reviewers will expect you to know.

The idea is not new or fringe. For two decades, physicists have shown that these methods separate signal from background better than hand-tuned cuts, and the field built standard tools around them. We do not invent statistics or hand-wave the hard parts. We show you the method, show you the evidence, and show you exactly how to check your own work before anyone else does.

Start with the free module

Get Module 1 and the one-page Roadmap — the exact path from where you are now to a finished, defensible analysis. No payment, no pressure.

Get Module 1 + the Roadmap →
Scientific references (20)
  1. Roe, B. P., Yang, H.-J., Zhu, J., Liu, Y., Stancu, I., & McGregor, G. (2005). Boosted decision trees as an alternative to artificial neural networks for particle identification. Nuclear Instruments and Methods in Physics Research A, 543(2-3), 577–584.
  2. Hoecker, A. et al. (2007). TMVA — Toolkit for Multivariate Data Analysis. arXiv:physics/0703039.
  3. Chen, T. & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794.
  4. Freund, Y. & Schapire, R. E. (1997). A decision-theoretic generalization of on-line learning and an application to boosting. Journal of Computer and System Sciences, 55(1), 119–139.
  5. Adam-Bourdarios, C., Cowan, G., Germain, C., Guyon, I., Kégl, B., & Rousseau, D. (2015). The Higgs boson machine learning challenge. JMLR Workshop and Conference Proceedings, 42, 19–55.
  6. Baldi, P., Sadowski, P., & Whiteson, D. (2014). Searching for exotic particles in high-energy physics with deep learning. Nature Communications, 5, 4308.
  7. Guest, D., Cranmer, K., & Whiteson, D. (2018). Deep learning and its application to LHC physics. Annual Review of Nuclear and Particle Science, 68, 161–181.
  8. Radovic, A. et al. (2018). Machine learning at the energy and intensity frontiers of particle physics. Nature, 560, 41–48.
  9. de Oliveira, L., Kagan, M., Nachman, B., & Schwartz, M. D. (2016). Learning particle physics by example: location-aware generative adversarial networks for physics synthesis. Computing and Software for Big Science, 1(1), 4.
  10. Louppe, G., Kagan, M., & Cranmer, K. (2017). Learning to pivot with adversarial networks. Advances in Neural Information Processing Systems, 30.
  11. Cranmer, K., Brehmer, J., & Louppe, G. (2020). The frontier of simulation-based inference. Proceedings of the National Academy of Sciences, 117(48), 30055–30062.
  12. Larkoski, A. J., Moult, I., & Nachman, B. (2020). Jet substructure at the Large Hadron Collider: a review of recent advances in theory and machine learning. Physics Reports, 841, 1–63.
  13. Butter, A. et al. (2022). Machine learning and LHC event generation. SciPost Physics, 14, 079.
  14. Nachman, B. (2020). A guide for deploying Deep Learning in LHC searches: how to achieve optimality and account for uncertainty. SciPost Physics, 8, 090.
  15. CMS Collaboration. (2020). A deep neural network to search for new long-lived particles decaying to jets. Machine Learning: Science and Technology, 1, 035012.
  16. ATLAS Collaboration. (2019). Performance of top-quark and W-boson tagging with ATLAS in Run 2 of the LHC. European Physical Journal C, 79(5), 375.
  17. Kasieczka, G. et al. (2019). The Machine Learning landscape of top taggers. SciPost Physics, 7, 014.
  18. Metodiev, E. M., Nachman, B., & Thaler, J. (2017). Classification without labels: learning from mixed samples in high energy physics. Journal of High Energy Physics, 2017(10), 174.
  19. Komiske, P. T., Metodiev, E. M., & Thaler, J. (2019). Energy flow networks: deep sets for particle jets. Journal of High Energy Physics, 2019(1), 121.
  20. Andreassen, A., Nachman, B., & Shih, D. (2020). Simulation assisted likelihood-free anomaly detection. Physical Review D, 101(9), 095004.

30-day full refund if the course does not meet your expectations. Try the full course. If it does not improve how you work, email us within 30 days and we will refund you — see our guarantee terms →

Ready to add machine learning to your analysis?

Start free. Get Module 1 and the one-page Roadmap — the exact path from where you are now to a result you can defend in review.

Get Module 1 + the Roadmap — Free →