Physics of the B-factories, MVA chapter 12nd workshop, KEK, May 2010 José Ocariz LPNHE – IN2P3 et Université Paris Diderot On behalf of the MVA chapter.

Physics of the B-factories, MVA chapter 1 2nd workshop, KEK, May 2010

José OcarizJosé Ocariz

LPNHE – IN2P3 et UniversitLPNHE – IN2P3 et Universitéé Paris Diderot Paris Diderot

On behalf of the MVA chapter editors and contributorsOn behalf of the MVA chapter editors and contributors

Multivariate Analysis :optimisation → tagging, PID, BKG suppression

1st “large” report

Physics of the B-factories, MVA chapter 2 2nd workshop, KEK, May 2010

Outline

Summary of decisions taken during 1st workshop at SLAC

Reorganisation of the former “MVA/Analysis Optimisation” chapters

Status of various subchapters

General Introduction / Analysis OptimisationTagging algorithmsPID algorithmsContinuum-suppresion algorithms

Manpower, timing issues

Physics of the B-factories, MVA chapter 3 2nd workshop, KEK, May 2010

Outcome from 1st worksop : reorganisation

From the 1st “strawman draft” :

a few disconnected chaptersflavour taggingmultivariate discriminantsanalysis optimisation

all refer to tools/methods based on similar methods

about 50 pages in total in 1st count

yet PID missing from listnot suitable for detector part

Physics of the B-factories, MVA chapter 4 2nd workshop, KEK, May 2010

Outcome from 1st worksop : reorganisation

(retained) reorganisation proposal :

a single chapter, with 4 sectionsa general introduction :

structures the overall landscapeuseful for students

3 specific applications :tagging, PID, bkg suppression

easier to manage (and decrease page length)

Physics of the B-factories, MVA chapter 5 2nd workshop, KEK, May 2010

Analysis optimization

SectionAnalysis optimizationSubsection

IntroductionNotationsFigures of meritMethodsSubsubsection

Rectangular cuts Linear discriminants

Neural netsBinary decision treesBoostingBaggingRandom Forest

Subsection Available tools Main section editor

Frank PorterIdentified contributors

Ilya Narski, José OcarizBelle ?

Main features :Pedagogical introductionFix notations

framework reference for thespecific applications following

± 6-8 pages

Physics of the B-factories, MVA chapter 6 2nd workshop, KEK, May 2010

Flavor tagging

SectionFlavor taggingSubsection

IntroductionDefinitionsPhysics souces of tagging informationMultivariate methods usedThings we tried, what worked, what didn’t ?

SubsectionFlavor tagging algorithms

BaBar and Belle algorithmsCategories

SubsectionMeasuring flavor tagging performance

History of tagging performance in Belle and BaBar Subsection

Systematic effectsTag-side interference

Main section editor(s)Juerg BeringerBelle ?

Main features :Example of MVA applicationSpecific to B-factories To be broadly referenced

± 8 pages

Physics of the B-factories, MVA chapter 7 2nd workshop, KEK, May 2010

Particle Identification

Main section editor(s) : TBD!Potential contributorsongoing contacts (BaBar), Belle ?

SectionParticle IdentificationSubsection

IntroductionDefinitionsDetector and physics souces of PIDMultivariate methods used

SubsectionPID algorithms

BaBar and Belle algorithms Subsection

Measuring PID performanceHistory of PID performance in Belle and BaBar

SubsectionSystematic effects

Main features :Example of MVA applicationLarge references to detector partDoes not fit into detector partTo be broadly referenced

± 6-8 pages ?

Outline could look like :

Physics of the B-factories, MVA chapter 8 2nd workshop, KEK, May 2010

Background suppression

SectionBackground suppression

SubsectionIntroductionSources of backgrounds: continuum, B-relatedSubsectionKinematical discrimination

mES/Mbc, E, intermediate resonant masses, angular distributions

SubsectionTopological discrimination

Information from complete event and/or ROEThrust, B momentum, energy flow, Legendre, SFW, etc…

SubsectionOptimisation criteria

S/B, significance, minimize systematics, CPU-related.linear vs. non-linear, treatment of correlations, etc…

SubsectionHistory SubsectionValidations, systematics

Main features :Various examples of MVA appsTo be broadly referenced

Feedback/interplay with physics chapters

Main section editorsHidekazu Kakuno, José Ocariz

Identified contributorsseveral volunteers (BaBar) ± 8 pages

Physics of the B-factories, MVA chapter 9 2nd workshop, KEK, May 2010

MVA-related topics

Discussion of some interesting MVA-related topics (e.g. time-evolution of performances, of optimisation criteria)would add value to the pedagogical purpose of the bookimplementation foreseen in the tagging and PID sections for bkg suppression, would be better illustrated with physics examples

Example from BaBar : optimisation criteria in B→,Ksuppress continuum vs. minimize systematicsdifferent requirements for CP analyses vs. BR analyses

could be discussed in the charmless chapters or here

Example from BaBar : background suppression in charmless Dalitz B→hhhtreatment of non-linear correlations wrt Dalitz variables“external” technical considerations (i.e. CPU)could be discussed in the charmless, Dalitz chapters or here

Decisions on whether/where to include such topics :requires discussion with global and physics editorsmay bring forward-referencings and (mild )revisiting of section structureneeds to check relevance on the Belle side

Physics of the B-factories, MVA chapter 10 2nd workshop, KEK, May 2010

Structure, manpower, timing issues

Overall chapter/section structuredesigned to ensure global coherence and minimize redundanciessuits the pedagogical purpose of the book

Manpowertagging, background-suppression ±OK …… less true for PID…

a few important contacts yet to be established need both BaBar/Belle contributors on all sections

→ homework for us chapter editors!

Timingmost chapter writing could start nowinterplay/feedback with physics chapters

check for specific needs and redundancieschannel-oriented illustrations of specific MVA features