MD High Energy Physics (HEP)

Purpose:
applications of modern statistical methods to data and models of high energy physics.

Background & Motivation:
High energy physics (HEP) experiments produce enormous quantities of data. Understanding and analysing this requires the use of well calibrated models of the underlying physical process alongside sophisticated data processing, extraction and manipulation methods. Historically analysis of HEP data has focused on inference for single observables, such as a particle mass or the value of a given constant, over more global analyses. There is a clear opportunity for application of MD methods of data handling, estimation and model approximation to begin new global analyses of many observables over a vast parameter space.

This will lead to better constraints on our understanding of the fundamental aspects of particle physics as well as revealing hitherto unknown correlations.

Progress in this direction will require the calibration and bias quantification of large scale HEP models, such as event generators and hydrodynamical simulations, along with the development of novel strategies for observable selection and model to data comparisions which can be carried out in finite computational time.

Topics to be considered:
Analysis of particle collision data from the LHC, understanding the role of computer models in the current HEP discovery process.

Applications of machine learning + data mining, pattern recognition and inference on large scale HEP data sets, how to sift out possibly interesting signals from mountains of noise.

Quantifying (confirmation (?)) biases in HEP data analysis arising from single observable focus, methods for estimation of systematic errors and of the power of these analyses.

Parameter estimation, observable selection and correlation estimation for very large scale models, including many parameters and many possible observables.

Aspects of model to data comparisons for very large multi-dimensional data sets: efficient implementations and approximations.

Methods for model to model comparison with models based on very different physics.