ml4hep

INFO: https://agenda.infn.it/event/28573/

This is a 2 days Training Course on Machine Learning for beginners focussed on High Energy Physics applications.

The course consists in theory (slides can be found here) and hands-on sessions and covers the following topics:

Day 1

• Bias/Variance in Machine Learning
• Gradient Descent (hands-on OGD, SGD, NAG, ADAM, RMSProp)
• Linear and Logistic Regression
• Combination of Models (Ensembles, Bagging, Boosting, Random Forests, GBT, XGBoost) (hands-on XGBoost Ex. 1, Ex. 2)
• Clustering (K-Means, Density-based clustering methods: DBSCAN, HDBSCAN) (hands-on Clustering)

Day 2:

• Introduction to Neural Networks and hyperparameter optimization (hands-on how to build your first feed-forward NN with PyTorch)
• Detector Design Optimization: single and multi-objective optimization (hands-on Bayesian Optimization, Multi-objective Optimization with meta-heuristic Ex. 2, MOO Ex. 3 (generalization to 3 objectives))

Credits/References:

The course utilizes the following as main reference:

[1] A high bias, low-variance introduction to Machine Learning [hblvi2ML]: https://arxiv.org/abs/1803.08823

Other references:

[2] Deep Learning, Ian Goodfellow, Yoshua Bengio and Aaron Courville, https://www.deeplearningbook.org

[3] Information Theory, Inference, and Learning Algorithms, David J.C. MacKay, https://www.inference.org.uk/itprnn/book.pdf

[4] AI4NP winter school, Detector design optimization, Cristiano Fanelli, https://github.com/cfteach/AI4NP_detector_opt

Requirements

python3; all other packages are installed from scratch during the course.

Documentation on scikit-learn can be found here.

Documentation on PyTorch can be found here.

We make use of jupyter notebook and colab.