root in particle physics - indico

Will Kalderon / 13 ROOT introduction

ROOT in particle physics

HASCO Summer School, 20.07.17

Will Kalderon, Lund University (with 99% of the credit to Olaf

Nackenhorst, University of Geneva)

1


Hooray!

2

• You finished the exams!

• Now: I will show you one of the main ways we actually do things in particle physics

• In a fairly short while: last evening in Göttingen, so I will try to be quick!


• Most of the material is from last year’s HASCO lecturer, Olaf Nackenhorst - enormous thanks to him, and all mistakes are down to me!

Introduction

3

FACULTÉ DES SCIENCESSection xxx

Olaf NackenhorstHASCO Summer School, 21.07.2016

Section xxx

2

Disclaimer & References

● Disclaimer

– Not a real lecture, rather an introductory tutorial / workshop

– Not complete, biased by personal point of view / experience

– Not a ROOT expert, just an experienced user

– Designed to match different experience levels

– Mainly practical examples → learning by doing using Jupyter Notebook

● References

– Parts based on last year's ROOT lecture by Andrea Knue

– Inspired by ROOT primer (Guide for beginners)

– ROOT web site: www.root.cern.ch

● THE source of information and help for all users (beginner & experts)

● User's Guide & Reference Guide (Documentation of classes)

● Download & installation instructions

● Topical manuals, tutorials, presentations, forum and more!

Olaf’s lecture

Andrea lecture

ROOT primer

root.cern.ch

User's guide

Reference guide

Download and install

https://indico.cern.ch/event/508168/contributions/2028731/

https://indico.cern.ch/event/388801/contributions/1822121/attachments/1131047/1616568/RootLecture.pdf

https://root.cern.ch/root/htmldoc/guides/primer/ROOTPrimer.pdf

http://www.root.cern.ch

https://root.cern.ch/root/htmldoc/guides/users-guide/ROOTUsersGuide.html

https://root.cern.ch/root/html606/

https://root.cern.ch/downloading-root


• Software framework for data processing, storage, analysis and visualisation

• Mainly written in C++, with bindings for python (“pyroot”)

• Almost everything I do in my research involves ROOT • This is true of most research in high-energy physics,

also used in astrophysics and neuroscience • LHC data is saved as ROOT files, processed using

ROOT, calibrated using ROOT, analysed using ROOT, and the plots for publication are made with ROOT

What is ROOT?

4


LOTS of data

5

• This file: 13 GB, from 1 minute of ATLAS running a few weeks ago • Lots of information to

store per event • 215,000 jets! • ~1/4-1/3 of the year =

1-200,000 minutes • Record data at 1kHz ->

PB of data, billions of events

ROOT (along with grid computing) allows us

to handle this

6



Section xxx

4

What we do in HEP (simplified)

Record data, reconstruct and store! Visualize data & publish!

+

Process data, select and analyze!

Phys. Rev. D 91, 012006 (2015)

Phys. Rev. D 91, 012006 (2015)

7



Section xxx

5

Standard tasks in HEP

● Standard Tasks in High Energy Physics

– Comparison of measurements to theoretical models

– Visualization of data

– Manipulation of data

– Fitting of data

– Statistical interpretation

– Storage of large data

– ...

● For all this and more: ROOT

– The following is based on ROOT6

– Jupyter Notebooks: > ROOT 6.05

● Visualise & share live code

● Web: http://jupyter.org/

8



Section xxx

6

Visualization Examples

9



Section xxx

7

Compiled vs Interactive ROOT

● Compiled ROOT

– Programming framework, not an office suite

– Write a program

– Compile as standalone application

● Interactive ROOT

– Comes with a C++ interpreter (Cling/ACLiC)

– Interactive C++ without the need of a compiler (like python)

– Just in time compilation – non trivial in C++ !

– Called ROOT prompt (interactive shell)

– Can interpret macros (non compiled programs)

10



Section xxx

8

Interactive ROOT - First Steps

● Type root to start

– root -l: no logo

– root -b: no graphics

– root -h: help (all options)

● Type .help or .?

– .L Plot.C: load macro

● Execute by Plot();

– .X Plot.C: execute macro

● Plot() needs to be defined!

● Compile a macro (ACLiC)

– root .L / .X Plot.C+

● Access shell command

– .!<command>

● Quit root: .q or .qqqq (force)

11



Section xxx

9

The TBrowser

● Load a root file during start

– root filename.root

● Inspect file with GUI

– new TBrowser

● Browse the file

– TTree::tree● Datacontainer

– TBranch::invariantMass● Content

● Draw content

– Click on the branch

– Use TTree::Draw● Selection & options

12



Section xxx

10

The FitPanel

● Open the FitPanel

– Right click on graph → FitPanel

Don't forget to save! Even as a root macro!

Nice, but write a script if you do something more than once!


Demo 1

13

Root basics

14



Section xxx

12

Example: Good vs bad plot

● Relatively easy to make plots in ROOT (left)

● Relatively difficult & time consuming to make nice plots (right)

15



Section xxx

13

The ROOT Color Wheel

● TColorWheel: 216 colors as used in web applications

● Special identifiers for colors

– kWhite, kBlack, kGrey

– kRed, kBlue, kGreen

– KYellow, kMagenta, kCyan

– And more…

● Colors in between

– Obtained by ± [1,10]

● Colorize objects

– SetLineColor()

– SetFillColor()

● Use colors smartly (grayscale)

You can also define any colour you like with RGB or hex

Sadly the Americans won the fight over spelling rights

16



Section xxx

14

Other useful attributes

● Marker style

– KDot, kPlus, kStar, kCircle, etc.

– SetMarkerStyle()

● Line style

– Fixed: 1-10

– Customizable: SetLineStyleString()

– SetLineStyle()

● Fill area style

– Fixed: 3000 + 1-25

– Customizable: FillStyle = 3ijk

● i=space between each hatch

● j=angle(<90), k=angle(>90)

– SetFillStyle()


Demo 2

17

Making pretty plots


Demo 2

18

Nice plots



Section xxx

16

Advanced Root Tools

● TMVA

– Toolkit for mulitvariate analysis

– Machine learning methods for classification and regression

– Documentation: Users guide and web site

– Exercise in tomorrow's tutorial!

● RooFit

– Toolkit for data modelling

– Maximum likelihood fitting, toy MC generator, visualisation

– Documentation: Quick introduction, RooFit in 20 min & users manual

● PROOF

– Parallel ROOT facility

– Parallelize running on large sets of ROOT files

– Documentation: Introduction

19



Section xxx

17

TMVA – Multivariate analysis

● Combine multiple inputs into a single (few) output

● Same processing for all methods, parallel run, easy to use, fair comparison

● Many machine learning techniques available (more coming)

– Boosted decision tree (BDT)

– Neural networks (NN)

– Probability density estimation (PDE)

– Function discriminant analysis (FDA)

– Support Vector Machine (SVM)

– Predictive learning (RuleFit)

– etc.

● Many visualisations

● Many monitoring tools

● Modern ML not yet integrated

20



Section xxx

18

PyROOT

● ROOT interface to Python

● Real mix of the two languages

● Power of C++

– compiled libraries

● Flexibility of Python

– Easy to use

● Introduction

– Tomorrow's tutorial!

xkcd.com


Conclusions

21

• ROOT is a powerful tool for particle physics • Most day-to-day work uses it to some degree • Many plotting options -> can be time-consuming to get it

just right • Different options for different use cases:

• Compiled C++ for speed (large-scale data processing) • pyroot for ease of use (plotting scripts) • interactive for quick checks

• I covered a tiny fraction of usage today - but there are lots of tutorials and lots of documentation out there!


A final word

22

• ROOT does a *lot* • Researchers at

CERN who have used it for many years still find new features

• Tomorrow I will help you through a taster • Rooms SR1 and SR2

• Google is your friend!

root in particle physics - indico

Documents