The topic of this course will be deep learning applications to high energy physics. I will focus primarily on collider physics and the LHC, but if time permits I may also broaden the focus to include other subfields relevant to HEP.

The last few years have seen the deep learning revolution come to HEP and the LHC. New algorithms and new approaches are being proposed to solve a wide array of problems, including jet tagging, event classification, pileup reduction, event generation, calorimeter simulation, and anomaly detection. The goal of this course is to provide an overview of these various deep learning applications at the LHC, with an eye towards current trends in research. Along the way, students will receive a practical introduction to fundamental concepts and methods in machine learning and data science. Hopefully, after taking this semester-long course, students will be able to understand the latest ML4HEP results on the arXiv and possibly even embark on research projects of their own.

**Instructor:**

Prof. David Shih

Office: Serin E370

Email:
dshih@physics.rutgers.edu

Phone: (848) 445-9072

Office hours: By appointment

**Lectures:** Wednesday 10:20-11:40AM and Friday 3:20-4:40PM, VIRTUAL

**Resources**:

- The Deep Learning Book. Goodfellow, Bengio and Courville (2016).

**Prerequisites**:

You will be expected to be at least somewhat familiar with python and some of its essential packages (numpy, matplotlib, ...). A
background in collider physics would also be helpful, although I will attempt to review the basics in the first few lectures.

You will NOT be required to know ROOT or any physics simulation tools (such as Madgraph, Pythia and Delphes). You will also NOT be required to know in advance any of the deep learning software frameworks such as Keras, Tensorflow and pytorch.

**Homeworks:**
TBD

**Exams:**
TBD

**Students with disabilities:**
Please read here.

- https://towardsdatascience.com/understanding-the-bias-variance-tradeoff-165e6942b229 Overfitting, underfitting, and bias-variance tradeoff
- https://towardsdatascience.com/gradient-descent-algorithm-and-its-variants-10f652806a3 Good description of gradient descent algorithm
- https://www.cs.cornell.edu/courses/cs4780/2018fa/lectures/lecturenote07.html More on gradient descent (and vs. Newton's method)

- https://ruder.io/optimizing-gradient-descent/ Great description of different gradient descent algorithms
- https://brilliant.org/wiki/backpropagation/ Detailed derivation of the backprop algorithm

- http://people.stern.nyu.edu/churvich/Regress/Handouts/Chapt7.pdf I based my proof of the Neyman-Pearson Lemma heavily off of this

- Pedagogical lectures on collider physics
- Pedagogical lectures on jet substructure and collider physics
- Book-length introduction to jet substructure and collider physics
- Fastjet user manual -- introduction contains an excellent, brief summary of jet clustering algorithms; later sections have a lot of useful detailed information
- Original reference for N-subjettiness

- Notebook showing how to create centered jet images
- Top tagging classifier examples
- Community top tagging comparison -- an excellent resource for state-of-the-art in jet tagging with deep learning at the LHC