\documentclass[]{beamer} \setbeamertemplate{navigation symbols}{} \usepackage{beamerthemesplit} \useoutertheme{infolines} \usecolortheme{dolphin} %\usetheme{Warsaw} \usetheme{progressbar} \usecolortheme{progressbar} \usefonttheme{progressbar} \useoutertheme{progressbar} \useinnertheme{progressbar} \usepackage{graphicx} %\usepackage{amssymb,amsmath} \usepackage[latin1]{inputenc} \usepackage{amsmath} \usepackage[T1]{fontenc} \usepackage{hepparticles} \usepackage{hepnicenames} \usepackage{hepunits} \usepackage{iwona} \progressbaroptions{imagename=images/lhcb} %\usetheme{Boadilla} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \definecolor{mygreen}{cmyk}{0.82,0.11,1,0.25} \setbeamertemplate{blocks}[rounded][shadow=false] \addtobeamertemplate{block begin}{\pgfsetfillopacity{0.8}}{\pgfsetfillopacity{1}} \setbeamercolor{structure}{fg=mygreen} \setbeamercolor*{block title example}{fg=mygreen!50, bg= blue!10} \setbeamercolor*{block body example}{fg= blue, bg= blue!5} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %\beamersetuncovermixins{\opaqueness<1>{25}}{\opaqueness<2->{15}} \title{Update on $\tau \to \mu \mu \mu$ searches} \author{M.Chrzaszcz$^{1,2}$,N. Serra$^1$, } %\date{\today} \begin{document} { \institute{$^1$ Zurich, $^2$ Krakow} \setbeamertemplate{footline}{} \begin{frame} \logo{ \vspace{2 mm} \includegraphics[height=1cm,keepaspectratio]{images/ifj.png}~ \includegraphics[height=1cm,keepaspectratio]{images/uzh.jpg}} \titlepage \end{frame} } \institute{UZH,IFJ} %\section[Outline]{} %\begin{frame} %\tableofcontents %\end{frame} \begin{frame}\frametitle{MC samples} \begin{enumerate} \only<1>{ \item MC Samples; quite nice(mostly Krakow) \begin{itemize} \item All cool MC generator cuts. \item Signal - DONE \item Calibration channel - DONE \item $b\overline{b}$ bck - DONE $18.1 pb^{-1}$ \item $c\overline{c}$ bck - $50$ DONE, $2.6pb^{-1}$ \item $Ds\to \eta(\mu\mu \gamma)\mu\mu$ - DONE - $>5fb^{-1}$ \item $\tau \to p \mu \mu$ Hopefully not needed :) \item Last night all samples got into ntuples. \end{itemize} \item $cc$, $bb$ cross section fixed for now(we will update if we have measurement for $cc$). } \end{enumerate} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%55 \begin{frame}\frametitle{Normalization} \only<1> { \begin{center} \begin{tiny} \begin{columns} \column{2.5in} \begin{center} $D_s \to \phi(\mu\mu)\pi$ in data.\\ \includegraphics[scale=0.13]{Ds_Mass/Ds_mass_data.png} \\ \begin{itemize} \item mean = $1970.3 \pm 0.9 MeV$ \end{itemize} \end{center} \column{2.5in} \begin{center} $D_s \to \phi(\mu\mu)\pi$ in MC.\\ \includegraphics[scale=0.13]{Ds_Mass/D_mass_base.png}\\ \begin{itemize} \item mean = $1969.1 \pm 0.60 MeV$ \end{itemize} \end{center} \end{columns} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{columns} \column{2.5in} \begin{center} \begin{small} \begin{itemize} \item $m_{\tau \to 3\mu} = \dfrac{1970.3}{1969.1} \times 1777.7 =$\textcolor{blue}{$ 1778.8 \pm 1.1 MeV$} \end{itemize} {~} \\ In agreement with 2011. \end{small} \end{center} \column{2.5in} \begin{center} Fit $\tau \to \mu\mu\mu$ in MC. \\ \includegraphics[scale=0.11]{Ds_Mass/tau_mass_base.png}\\ % \begin{itemize} % \item mean = $1777.7 \pm 0.4 MeV$ \\ % \end{itemize} \end{center} \end{columns} \end{tiny} \end{center} } \end{frame} \begin{frame}\frametitle{Trigger} \begin{enumerate} \only<1>{ \item Here we really suck. \begin{itemize} \item Trigger lines changed between 2011 and 2012 \item In 2012 also lines have changed... \item Need to evaluate the efficiency for each TCK. \item I am preparing all possible ntuples for Jon to weight the efficiencies accordingly to TCK version. \item God have mercy on my soul... \end{itemize} } \end{enumerate} \end{frame} \begin{frame}\frametitle{DATA -MC comparison} {~} \only<1> { \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p0_IP.png} \\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p0_IPSig.png}\\ \end{columns} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p1_IP.png}\\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p1_IPSig.png}\\ \end{columns} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{$\PDs \to \eta(\mu \mu \gamma) \mu \nu$} \only<1> { \begin{exampleblock}{~} \begin{enumerate} \item The dominant background source of peaking background in this analysis is \textcolor{blue}{$\PDs \to \eta(\mu\mu\gamma) \mu \nu$}\\ \item In 2011 we suffered from lack of MC statistics. \item Thanks to generator cuts our pdfs became more stable. \item Pdf used: $\mathcal{P} = exp(m) \times Pol^n(m)$ \item This is ready to go. \end{enumerate} \end{exampleblock} \begin{columns} \column{2.5in} \begin{center} \includegraphics[scale=0.09]{RD_meeting/pid_0_65_0_725geo-0_48_0_05.png} \\ \begin{tiny} PID:$0.65;0.725$,GEO:$-0.48;0.05$ \end{tiny} \end{center} \column{2.5in} \begin{center} \includegraphics[scale=0.09]{RD_meeting/pid_0_725_0_86geo0_35_0_65.png}\\ \begin{tiny} PID:$0.725;0.0.86$,GEO:$0.35;0.65$ \end{tiny} \end{center} \end{columns} } % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{Isolating Parameter} {~} \only<1> { \begin{itemize} \item All the R\&D has finished. \item I have an optimum isolating parameter for 5 different tau sources. \item Only need to write a DV algorithm to put this inside zoontuple. \item Also needs comparison to iso and non -isolating.(Still didn't get answer when can this happen). \end{itemize} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{frame}\frametitle{MVA} {~} \only<1> { \begin{itemize} \item All the scripts are there \item Limitation is the $cc$ bck sample. Would be nice to have two times more. \item Let's hope this plot will stay the same :) \includegraphics[scale=0.27]{images/BDT_comparison.png} \end{itemize} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{frame}\frametitle{Binning optimisation} {~} \only<1> { \begin{itemize} \item Also done(I used 2011 data, so just when we fix new BDT need to press Enter). \item How ever last night I had an idea(Nico you won't like this one). What about use purelly Bayesian way to optimise? \item I am to curious to get discourage not to do it :) \end{itemize} \begin{columns} \column{2.5in} \includegraphics[scale=0.13]{inflaton/punzi1.png} \begin{itemize} \item FOM as a function of N. of bins. \end{itemize} \column{2.5in} \includegraphics[scale=0.27]{RD_meeting/2d-data.pdf} \begin{itemize} \item Signal efficiency in 2011 binning. \end{itemize} \end{columns} } \end{frame} \begin{frame}\frametitle{Model dependence} {~} \only<1> { \begin{itemize} \item Paul implemented an "model independent" 3 scenarios. \item he wants only to correct Normalization for studies. \item With Nico we think multidimensional fir would be more fun. \item Also would like to implement some SUSY models. \end{itemize} } \begin{center} \begin{columns} \column{1.6in} {~}$(\overline{L} \gamma_{\mu} L)(\overline{L} \gamma^{\mu} L)$\\ {~}\includegraphics[scale=0.22]{images/gammallll.png} \column{1.6in} $(\overline{R}\gamma_{\mu} R)(\overline{L} \gamma^{\mu} L)$\\ \includegraphics[scale=0.22]{images/gammallrr.png} \column{1.6in} $g'(\overline{L}H\sigma_{\mu\nu}R)B^{\mu\nu}$\\ \includegraphics[scale=0.22]{images/gammarad.png} \end{columns} \end{center} \end{frame} \begin{frame}\frametitle{Conclusions} {~} \only<1> { \begin{itemize} \item Analysis is well under way. \item I am determined to finish asap. \item End of this year is possible if we won't do $\tau \to p \mu \mu$. \end{itemize} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{frame}\frametitle{} \begin{Huge} BACKUP \end{Huge} \end{frame} \begin{frame}\frametitle{Status} \begin{columns} \column{2in} \begin{center} $1 $fb$^{-1}$ analysis of \textcolor{violet}{$\tau \to \mu \mu \mu$} and \textcolor{blue}{$\tau \to p \mu \mu$} appeared in PLB. \end{center} \column{3in} \includegraphics[scale=0.197]{RD_meeting/PLB.png} \end{columns} \begin{exampleblock}{2011 results:} \begin{enumerate} \item Obtained limit for $\tau \to \mu \mu \mu$: $8.0 \times 10^{-8}$. \item Belle(BaBar) results: $2.1 (3.2) \times 10^{-8}$ at $90\%$ CL. \item For 2012 + 2011 planned to implement several improvements. \end{enumerate} \end{exampleblock} % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{Status} For now we use: \begin{enumerate} \item Stripping 20. \item Signal sample: official+Krakow produced sample $(1M+1M)$. \item $bb$ and $cc$ samples: official+Krakow. In total 30M events. \item General strategy stays the same as 2011. \end{enumerate} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \section{MC Samples} \begin{frame}\frametitle{Cross section update} \only<1> { Analysis uses the knowledge of $c\overline{c}$ and $b\overline{b}$ cross sections. In 2011 both were measured by LHCb. For 2012 for the moment we assume: \begin{itemize} \item $\sigma_{b\overline{b}}^{8TeV}=298\pm36 \mu b$ from LHCB-PAPER-2013-016 \item $\sigma_{c\overline{c}}^{8TeV}=\sigma_{c\overline{c}}^{7TeV}\times \dfrac{8}{7} = 6950 \pm 1100 \mu b$ \end{itemize} \begin{exampleblock}{Cross checks on $c\overline{c}$} \begin{enumerate} \item Pythia cross section calculation. \item Comparing $D_s$ yields in data. \end{enumerate} \end{exampleblock} } %\textref {M.Chrz\k{a}szcz 2013} \end{frame} \begin{frame}\frametitle{Generated MC samples } \only<1> { \begin{enumerate} \item In the 2011 analysis one of the complications from MC was the wrong mixture of tau sources. \item For 2012 we solved this problem by simulating signal in 5 parts. One for each production channel: \end{enumerate} } \begin{center} \fcolorbox{blue}{yellow}{ %\begin{equation}NUmber of ne $\tau \to \mu \mu \mu = \begin{cases} \PB \to \Ptau \to \mu \mu \mu & 11.6\% \\ \PB \to \PDs \to \tau \to \mu \mu \mu & 8.7\% \\ \PB \to \PD \to \tau \to \mu \mu \mu & 0.2\% \\ \PDs \to \tau \to \mu \mu \mu & 75.0\% \\ \PD \to \tau \to \mu \mu \mu & 4.4\% \\ \end{cases}$ %\end{equation} } % $\HepParticle{B}{}{\pm} \to \HepParticle{D}{}{(\ast)} \tau^{\pm} \nu$} \end{center} %\textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%55 \begin{frame}\frametitle{MC Generator Cuts} \only<1> { In order to use computing resources in more efficient way we introduced generator level cuts. \begin{center} \begin{tabular}{ | c | c || l | c |} \hline \multicolumn{2}{|c|| }{Signal sample\footnote{$X \to \tau\to 3\mu$, $\PDs \to \eta(\mu\mu \gamma) \mu \nu$, $\PDs \to \phi(\mu\mu) \pi$ }} & \multicolumn{2}{|c| }{Background sample(Dimuon)\footnote{$c\bar{c}$, $b\bar{b}$ }} \\ \hline \hline $p_{t\mu}$ & $>250MeV$ & $p_{t\mu}$ & $>280MeV$ \\ \hline $p_{\mu}$ & $>2.5GeV$ & $p_{\mu}$ & $>2.9GeV$ \\ \cline{3-4} & & $m(\mu\mu)$ & $<4.5GeV$\\ \cline{3-4} & & DOCA$(\mu\mu)$ & $<0.35mm$\\ \hline \end{tabular} \end{center} } Gain a factor of $\sim 2-3$ in signal statistics compared to 2011 and factor of ~8 in background. %\textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{frame}\frametitle{Trigger lines} \only<1> { In 2011 we took all trigger lines into account. Studies shown we can gain on limiting ourselves to specific lines (2011 data sample). \begin{center} \begin{tabular}{| c | c | c | c | c | } \hline Line Name & $\epsilon [\%]$ & $\epsilon' [\%]$ & $\beta [\%]$ & $\beta' [\%]$ \\ \hline \hline Hlt2CharmSemilepD2HMuMu & $81.7$ & $81.7$ & $56.8$ & $56.8$ \\ \hline Hlt2DiMuonDetached & $75.0$ & $12.5$ & $54.1$ & $17.6$ \\ \hline Hlt2TriMuonTau & $66.3$ & $2.9$ & $60.0$ & $12.2$ \\ \hline Others & - & $2.2$ & - & $11.6$ \\ \hline \end{tabular} \end{center} , where $\epsilon$ is the signal efficiency (any Hlt2physics), $\epsilon'$ is the gain of the efficiency.\\ $\beta$ is the efficiency of background and $\beta'$ is the gain of the bck efficiency\\ Rule of thumb (using $\frac{s}{\sqrt{b}}$ FOM) tells us that we can gain $\mathcal{O}(5\%)$. } %\textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%55 \section{Normalization} \begin{frame}\frametitle{Normalization channel} \only<1> { As last year we will use \textcolor{blue}{$\PDs \to \phi(\mu\mu) \pi$}.Similarly to signal channels we produced them with correct proportion: \begin{exampleblock}{~} \begin{enumerate} \item $cc \to \PDs \to \phi (\mu\mu) \pi$ $89.7\%$ \item $bb \to \PDs \to \phi (\mu\mu) \pi$ $10.3\%$ \end{enumerate} \end{exampleblock} We avoid reweighting of the samples as in 2011. } %\textref {M.Chrz\k{a}szcz 2013} \end{frame} \begin{frame}\frametitle{Mass correction} \only<1> { \begin{center} \begin{tiny} \begin{columns} \column{2.5in} \begin{center} $D_s \to \phi(\mu\mu)\pi$ in data.\\ \includegraphics[scale=0.13]{Ds_Mass/Ds_mass_data.png} \\ \begin{itemize} \item mean = $1970.3 \pm 0.9 MeV$ \end{itemize} \end{center} \column{2.5in} \begin{center} $D_s \to \phi(\mu\mu)\pi$ in MC.\\ \includegraphics[scale=0.13]{Ds_Mass/D_mass_base.png}\\ \begin{itemize} \item mean = $1969.1 \pm 0.60 MeV$ \end{itemize} \end{center} \end{columns} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{columns} \column{2.5in} \begin{center} \begin{small} \begin{itemize} \item $m_{\tau \to 3\mu} = \dfrac{1970.3}{1969.1} \times 1777.7 =$\textcolor{blue}{$ 1778.8 \pm 1.1 MeV$} \end{itemize} {~} \\ In agreement with 2011. \end{small} \end{center} \column{2.5in} \begin{center} Fit $\tau \to \mu\mu\mu$ in MC. \\ \includegraphics[scale=0.11]{Ds_Mass/tau_mass_base.png}\\ % \begin{itemize} % \item mean = $1777.7 \pm 0.4 MeV$ \\ % \end{itemize} \end{center} \end{columns} \end{tiny} \end{center} } % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{Background samples normalization} \only<1> { For the normalization of background samples($c\bar{c}$ and $b\bar{b}$) we used generator cuts efficiencies and corrected the nominal cross section accordingly:\\ \begin{center} $\mathcal{L} = \dfrac{N_{MC}}{\varepsilon_{acc} \times \varepsilon_{gen} \times \sigma_{LHCb}}$ \end{center} The obtained luminosities(per 1M events): \begin{exampleblock}{~} \begin{enumerate} \item $\mathcal{L}_{cc} = 0.25 \pm 0.04 pb^{-1}$ \item $\mathcal{L}_{bb} = 1.20 \pm 0.15 pb^{-1}$ \end{enumerate} \end{exampleblock} } Dominant uncertainty from the cross section. % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \section{Peaking backgrounds} \begin{frame}\frametitle{$\PDs \to \eta(\mu \mu \gamma) \mu \nu$} \only<1> { \begin{exampleblock}{~} \begin{enumerate} \item The dominant background source of peaking background in this analysis is \textcolor{blue}{$\PDs \to \eta(\mu\mu\gamma) \mu \nu$}\\ \item In 2011 we suffered from lack of MC statistics. \item Thanks to generator cuts our pdfs became more stable. \item Pdf used: $\mathcal{P} = exp(m) \times Pol^n(m)$ \end{enumerate} \end{exampleblock} \begin{columns} \column{2.5in} \begin{center} \includegraphics[scale=0.09]{RD_meeting/pid_0_65_0_725geo-0_48_0_05.png} \\ \begin{tiny} PID:$0.65;0.725$,GEO:$-0.48;0.05$ \end{tiny} \end{center} \column{2.5in} \begin{center} \includegraphics[scale=0.09]{RD_meeting/pid_0_725_0_86geo0_35_0_65.png}\\ \begin{tiny} PID:$0.725;0.0.86$,GEO:$0.35;0.65$ \end{tiny} \end{center} \end{columns} } % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{frame}\frametitle{$D \to \Ph \Ph \Ph $} \only<1> { In 2011 we saw a triple miss-ID background: $\PDp \to \PK \Ppi \Ppi$. This background was in trash-bins that were not used in the analysis.\\ Also new sources of bck($D_x\to 3\pi$) are well under control. \begin{columns} \column{1.6in} \begin{center} \includegraphics[scale=0.17]{images/pipipi_peak_2011.pdf}\\ \begin{itemize} \item 2011 data \end{itemize} \end{center} \column{1.6in} \begin{center} \includegraphics[scale=0.17]{images/pipipi_peak_2012.pdf}\\ \begin{itemize} \item 2012 data \end{itemize} \end{center} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%55 \column{1.6in} \begin{center} \includegraphics[scale=0.19]{images/FittoDkpipi_2012.pdf}\\{~}\\ \begin{itemize} \item 2012 data \end{itemize} \end{center} \end{columns} {~}\\ In 2012 there is still no significant amount of triple mis-ID background in the bins important to the analysis. } % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \section{MVA development} \begin{frame}\frametitle{Isolating parameters} {~} \only<1> { Inputs for isolating parameter(based on Giampiero work): \begin{center} \begin{tabular}{ | c | p{10cm} |} \hline Variable & Description \\ \hline \hline IP $\chi^2$ & Impact parameter $\chi^2$ wrt any PV \\ \hline IP & Impact parameter wrt any PV \\ \hline angle & angle between $\mu$ and track \\ \hline doca & doca between the $\mu$ and the track \\ \hline PVdis & $\vert \overrightarrow{TV} - \overrightarrow{PV} \vert$, signed according to $z_{TV} - z_{PV}$. \\ \hline SVdis & $\vert \overrightarrow{TV} - \overrightarrow{SV} \vert$, signed according to $z_{STV} - z_{PV}$. \\ \hline fc & $\dfrac{\vert \overrightarrow{P_{\mu}} + \overrightarrow{P_{tr}} \times \alpha }{\vert \overrightarrow{P_{\mu}} + \overrightarrow{P_{tr}} \times \alpha + P_{T_{\mu}}+ P_{T_{tr}}}$\footnote{$\alpha$ is the angle between $\overrightarrow{P}_{\mu} + \overrightarrow{P}_{tr}$ and $\overrightarrow{PV} - \overrightarrow{TV}$ }\\ \hline \end{tabular} \end{center} } \only<2> { \begin{enumerate} \item In 2011 we used the isolation parameter developed for $\PBs \to \mu\mu$. For 2012 data we optimised the isolation parameter for our channel based on MVA(BDT). \item We follow two approaches: train a MVA on signal vs. bkg tracks, and the isolating vs. non-isolating tracks. \item We see a big improvement compared to old isolation. \end{enumerate} \begin{columns} \column{1.6in} \begin{center} \includegraphics[scale=0.15]{RD_meeting/mva_BDT.png} \\ \end{center} \column{1.6in} \begin{center} \includegraphics[scale=0.15]{RD_meeting/rejBvsS.png}\\ \end{center} \column{1.6in} \begin{center} \includegraphics[scale=0.15]{images/Laura/rejBvsS.png}\\ \end{center} \end{columns} } % \textref {M.Chrz\k{a}szcz 2013} \end{frame} \begin{frame}\frametitle{Ensemble Selection} {~} \begin{exampleblock}{~} \begin{enumerate} \item In the last few years people winning leading machine learning contests started to combine their classifiers to squeeze the best out of them. \item This technique/method is know as Ensemble Selection or Blending. \item The plan for $\tau \to \mu \mu \mu$ is to take it to the next level. \item Combine not only different signal classifiers, but also different $\tau$ sources(slide 4). \item Allows for usage different isolating parameters for each channel. \end{enumerate} \end{exampleblock} \end{frame} \begin{frame}\frametitle{Ensemble Selection - How to} {~} How to make an Ensemble Selection \begin{exampleblock}{~} \begin{enumerate} \item Construct a reduced training set. \item Train you different models on the reduced training set. \item Combine/Blend all the models on the rest of the data set. \item The output is a function that mixes the individual model predictions into a blended prediction, hopefully better than any individual result. \end{enumerate} \end{exampleblock} \end{frame} \begin{frame}\frametitle{Ensemble Selection} %%%%%%%%%%%%%%%%%%%%%%%5 \begin{columns} \column{1.6in} \begin{center} \includegraphics[scale=0.15]{RD_meeting/rejBvsS_21513000.png}\\ \begin{itemize} \item $\PB \to \PD \to \tau$ \end{itemize} \end{center} \column{1.6in} \begin{center} \includegraphics[scale=0.15]{RD_meeting/rejBvsS_21513001.png}\\ \begin{itemize} \item $\PD \to \tau$ \end{itemize} \end{center} \column{1.6in} \begin{center} \includegraphics[scale=0.15]{RD_meeting/rejBvsS_23513000.png}\\ \begin{itemize} \item $\PB \to \PDs \to \tau$ \end{itemize} \end{center} %\column{2.5in} %\begin{center} % \includegraphics[scale=0.15]{RD_meeting/rejBvsS_23513001.png}\\ % \begin{itemize} % \item $\PDs \to \tau$ % \end{itemize} %\end{center} \end{columns} \end{frame} \begin{frame}\frametitle{Ensemble Selection} % \begin{columns} %\column{2.5in} % \includegraphics[scale=0.2]{RD_meeting/rejBvsS_oryginal.png} % \column{2.5in} % \includegraphics[scale=0.2]{RD_meeting/rejBvsS_blend.png} % \end{columns} \begin{center} \includegraphics[scale=0.3]{images/BDT_comparison.png} \end{center} % \textref {M.Chrz\k{a}szcz, N.Serra 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5555 \section{Binning optimisation} \begin{frame}\frametitle{Binning optimisation} {~} \only<1> { For the 2011 analysis we had two classifiers: $PIDNN$ and $M_{GEO}$. Each of them we optimised separately. For the 2012 analysis we are performing a simultaneous 2D optimisation. \begin{columns} \column{2.5in} \includegraphics[scale=0.13]{inflaton/punzi1.png} \begin{itemize} \item FOM as a function of N. of bins. \end{itemize} \column{2.5in} \includegraphics[scale=0.27]{RD_meeting/2d-data.pdf} \begin{itemize} \item Signal efficiency in 2011 binning. \end{itemize} \end{columns} } \end{frame} \section{Model dependence} \begin{frame}\frametitle{Model dependence} \begin{exampleblock}{Minimal Lepton Flavour Violation Model\footnote{arXiv:0707.0988}} \begin{itemize} \item In effective-field-theory we introduce new operators that at electro-weak scale are compatible with $SU(2)_L \times U(1)$. \item Left handed lepton doublets add right handed lepton singlets follow the group symmetry: $G_{LF} = SU(3)_L \times SU(3)_E$. \item LFV arises from breaking this group. \item We focus on three operators that have dominant contribution to NP: \begin{enumerate} \item Purely left handed iterations: $(\overline{L} \gamma_{\mu} L)(\overline{L} \gamma^{\mu} L)$ \item Mix term: $(\overline{R}\gamma_{\mu} R)(\overline{L} \gamma^{\mu} L)$ \item Radiative operator: $g'(\overline{L}H\sigma_{\mu\nu}R)B^{\mu\nu}$ \end{enumerate} \end{itemize} \end{exampleblock} \end{frame} \begin{frame}\frametitle{Reweighting MC samples} \only<1>{ \begin{center} \begin{columns} \column{2.5in} {~}Reconstruction:\\ {~}\includegraphics[scale=0.22]{images/acceptance.png} \column{2.5in} Offline:\\ \includegraphics[scale=0.22]{images/offline.png} \end{columns} \end{center} } \only<2>{ \begin{center} \begin{columns} \column{1.6in} {~}$(\overline{L} \gamma_{\mu} L)(\overline{L} \gamma^{\mu} L)$\\ {~}\includegraphics[scale=0.22]{images/gammallll.png} \column{1.6in} $(\overline{R}\gamma_{\mu} R)(\overline{L} \gamma^{\mu} L)$\\ \includegraphics[scale=0.22]{images/gammallrr.png} \column{1.6in} $g'(\overline{L}H\sigma_{\mu\nu}R)B^{\mu\nu}$\\ \includegraphics[scale=0.22]{images/gammarad.png} \end{columns} \end{center} } \begin{equation} \epsilon_{gen\&rec} = C\epsilon^{LHCbMC}_{gen\&rec} \sum \rho^{model}(m_{12},m_{23}) \end{equation} \only<1>{ \begin{itemize} \item Simulated signal events with PHSP \item Take into account reconstruction and selection. \item Reweight accordingly to a given distribution. \end{itemize} } \only<2>{ \begin{itemize} \item Simulated signal events with PHSP \item Take into account reconstruction and selection. \item Reweight accordingly to a given distribution. \end{itemize} } \end{frame} \section{Conclusions} \begin{frame}\frametitle{Conclusions} {~} \only<1> { \begin{exampleblock}{~} \begin{enumerate} \item Analysis is well underway. \item More efficient use of computing resources and increased MC statistics helps at all ends \item Hope to improve the MVA/binning. \end{enumerate} \end{exampleblock} } \includegraphics[scale=0.4]{RD_meeting/phd052805.png}\\ \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{frame} {~} \begin{Huge} BACKUP \end{Huge} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%55 \begin{frame}\frametitle{$B \to \tau$} {~}\\ We really suck in selecting this channel. \includegraphics[scale=0.4]{tmva/ROC_31113002.png} % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{$B \to D_s \to \tau$} {~}\\ On the biggest contributing channel we are quite optimal. \includegraphics[scale=0.4]{tmva/ROC_23513000.png} % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{$D_s \to \tau$} {~}\\ On the biggest contributing channel we are quite optimal. \includegraphics[scale=0.4]{tmva/ROC_23513001.png} %\textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{$B \to D^+ \to \tau$} {~}\\ On the biggest contributing channel we are quite optimal. \includegraphics[scale=0.4]{tmva/21513000_roc2.png} % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{$D^+ \to \tau$} {~}\\ On the biggest contributing channel we are quite optimal. \includegraphics[scale=0.4]{tmva/ROC_21513001.png} %\textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{Comparison on mix sample} {~}\\ On the biggest contributing channel we are quite optimal. \includegraphics[scale=0.4]{tmva/mix.png} %\textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{Conclusions on TMVA} {~}\\ \begin{itemize} \item Each of the signal components is enormously larger than MVA trained on mix. \item Method looks very promising if we can find a nice blending method(work for next week). \item Mayby discusion on TMVA/MatrixNet/Neurobayes is next to leading order effect compared to this method? \end{itemize} % \textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5 \begin{frame}\frametitle{Comparison on mix sample} {~}\\ \begin{columns} \column{2.5in} \includegraphics[scale=0.27]{RD_meeting/rejBvsS_oryginal.png} \column{2.5in} \includegraphics[scale=0.27]{RD_meeting/rejBvsS_blend.png} \end{columns} %\textref {M.Chrz\k{a}szcz 2013} \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{frame}\frametitle{$D_s$ correction} {~} \only<1> { \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/cdf1.png} \\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/cdf2.png}\\ \end{columns} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/ciso.png}\\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/doca12.png}\\ \end{columns} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5555 \begin{frame}\frametitle{$D_s$ correction} {~} \only<1> { \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/doca23.png} \\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/doca13.png}\\ \end{columns} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/FD.png}\\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/FDE.png}\\ \end{columns} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5555 \begin{frame}\frametitle{$D_s$ correction} {~} \only<1> { \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/IP.png} \\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/isoa.png}\\ \end{columns} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/isob.png}\\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/isoc.png}\\ \end{columns} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5555 \begin{frame}\frametitle{$D_s$ correction} {~} \only<1> { \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/isod.png} \\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/isoe.png}\\ \end{columns} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/isof.png}\\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/Life_time.png}\\ \end{columns} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5555 \begin{frame}\frametitle{$D_s$ correction} {~} \only<1> { \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p0_IP.png} \\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p0_IPSig.png}\\ \end{columns} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p1_IP.png}\\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p1_IPSig.png}\\ \end{columns} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5555 \begin{frame}\frametitle{$D_s$ correction} {~} \only<1> { \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p2_IP.png} \\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/p2_IPSig.png}\\ \end{columns} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{columns} \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/pangle.png}\\ \column{2.5in} \includegraphics[scale=0.18]{Ds_Splot/pt.png}\\ \end{columns} } \end{frame} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5555 \begin{frame}\frametitle{$D_s$ correction} {~} \only<1> { \includegraphics[scale=0.18]{Ds_Splot/vtxchi2.png} \\ } \end{frame} \end{document}