The main MLHEP program will cover the following topics
- Classification / Regression models
- Algorithm composition, boosting
- Feature selection / Dimensionality reduction
- Model quality criteria selection and model evaluation
- Neural Networks introduction
- Model overfitting detection and mitigation
- Hyper-parameter optimization for predictive models
- Decorrelation of variables and predictions
- Deep learning approach
Each lecture will be accompanied by a seminar, which will allow for students to gain practical experience and participate in group discussions. Students will have to use their own laptops to participate in the seminars.
For those participants who feel rather comfortable with the basics concepts of Machine Learning we provide advanced track, that will take place in parallel to the main track during Monday-Wednesday in Andromeda room (same building). The timeslots for lectures/seminars of the advanced track will coincide with the timeslots of the main track. The topics of the advanced track cover:
1) Trigger system overview, machine learning in the LHCb topological
trigger; speeding-up predictions for boosted decision trees and neural
networks. (Monday)
2) Tools and practices for reproducible research design and conduction (Tuesday)
3) Tracking approaches overview, methods and tools (Wednesday)
In addition to the Machine Learning lectures and seminars, a variety of talks by HEP-practitioners who used Machine Learning methods for solving particular problems in High Energy Physics.
During the MLHEP school our participants will have opportunity to join and compete in data challenge specially organized for this event.
