main.tex 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334
  1. % rubber: module pdftex
  2. \documentclass[english,aspectratio=43,8pt]{beamer}
  3. \usepackage{graphicx}
  4. \usepackage{amssymb}
  5. \usepackage{booktabs}
  6. \usepackage{siunitx}
  7. \usepackage{subcaption}
  8. \usepackage{marvosym}
  9. \usepackage{verbatim}
  10. \usepackage[normalem]{ulem} % Needed for /sout
  11. \newcommand{\pb}{\si{\pico\barn}}%
  12. \newcommand{\fb}{\si{\femto\barn}}%
  13. \newcommand{\invfb}{\si{\per\femto\barn}}
  14. \newcommand{\GeV}{\si{\giga\electronvolt}}
  15. \hypersetup{colorlinks=true,urlcolor=blue}
  16. \usetheme[]{bjeldbak}
  17. \newcommand{\backupbegin}{%
  18. \newcounter{finalframe}
  19. \setcounter{finalframe}{\value{framenumber}}
  20. }
  21. \newcommand{\backupend}{%
  22. \setcounter{framenumber}{\value{finalframe}}
  23. }
  24. \newcommand\blfootnote[1]{%
  25. \begingroup
  26. \renewcommand\thefootnote{}\footnote{#1}%
  27. \addtocounter{footnote}{-1}%
  28. \endgroup
  29. }
  30. \begin{document}
  31. \title[$e$ Seeding Validation]{Offline Electron Seeding Validation \-- Update}
  32. \author[C. Fangmeier]{\textbf{Caleb Fangmeier} \\ Ilya Kravchenko, Greg Snow}
  33. \institute[UNL]{University of Nebraska \-- Lincoln}
  34. \date{EGM Reco/Comm/HLT meeting | June 22, 2018}
  35. \titlegraphic{%
  36. \begin{figure}
  37. \includegraphics[width=1in]{CMSlogo.png}\hspace{0.75in}\includegraphics[width=1in]{nebraska-n.png}
  38. \end{figure}
  39. }
  40. \begin{frame}[plain]
  41. \titlepage%
  42. \end{frame}
  43. \begin{frame}{Introduction}
  44. \begin{itemize}
  45. \item Our goal is to study \textbf{seeding} for the \textbf{offline} GSF tracking with the \textbf{new pixel detector}.
  46. \item Specifically, we want to optimize the new pixel-matching scheme from HLT for use in off-line reconstruction.
  47. \item This Talk:
  48. \begin{itemize}
  49. \item Define and demonstrate performance of a GSF-Track ``Fake Rate'' for:
  50. \begin{itemize}
  51. \item Current offline (Legacy HLT) seeding method with default offline settings
  52. \item New seeding method with HLT settings\footnotemark%
  53. \item New seeding method with optimized-for-offline (aka \texttt{wide}) settings
  54. \end{itemize}
  55. \item Show efficiency for prompt electrons specifically
  56. \end{itemize}
  57. \end{itemize}
  58. \footnotetext[1]{\tiny Note: In previous talks I've called this one \texttt{narrow}.}
  59. \end{frame}
  60. \begin{frame}{N-Hit Electron Seeding}
  61. \begin{columns}
  62. \begin{column}{0.5\textwidth}
  63. {\small
  64. \begin{enumerate}
  65. \item Using the beam spot, the SC position, and SC energy, propagate a path through the pixels.
  66. \item Require the first hit to be within a $\delta\phi$ and $\delta z$ window. ($\delta\phi$ and $\delta R$ for FPIX)
  67. \item $\delta z$ window for first hit is huge as SC and beam spot positions give very little information about $z$.
  68. \item Forget the SC position, and propagate a new track based on the vertex and first hit positions, and the SC energy.
  69. \item Progress one-by-one through the remaining hits in the seed and require each one fit within a specified window around the track.
  70. \item Quit when all hits are matched, or a hit falls outside the window. No skipping is allowed.
  71. \end{enumerate}
  72. }
  73. \end{column}
  74. \begin{column}{0.5\textwidth}
  75. \begin{figure}
  76. \includegraphics[width=0.9\textwidth]{../common/diagrams/seeding_step2.png}
  77. \end{figure}
  78. \begin{figure}
  79. \includegraphics[width=0.9\textwidth]{../common/diagrams/seeding_step3.png}
  80. \end{figure}
  81. \end{column}
  82. \end{columns}
  83. \end{frame}
  84. \begin{frame}{Definitions}
  85. \begin{itemize}
  86. \item \textbf{Sim-Track \--} A track from a simulated electron both originating from the luminous region of CMS (beam-spot +- 5$\sigma$) and having $|\eta|<3.0$.
  87. \item \textbf{ECAL-Driven Seed \--} A seed created via a matching procedure between Super-Clusters and General Tracking Seeds (Either from \texttt{ElectronSeedProducer} or \texttt{ElectronNHitSeedProducer}). Must have $HOE<0.15$.
  88. \item \textbf{GSF Track \--} A track from GSF-Tracking resulting from an \textbf{ECAL-Driven Seed}
  89. % \item \textbf{Seeding Efficiency \--} The fraction of \textbf{Sim-Tracks} that have a matching \textbf{ECAL-Driven Seed} (based on simhit-rechit linkage or $\Delta R$ matching)
  90. \item \textbf{GSF Tracking Efficiency \--} The fraction of \textbf{Sim-Tracks} that have a matching \textbf{GSF Track} (based on $\Delta R$ matching)
  91. % \item \textbf{ECAL-Driven Seed Purity \--} The fraction of \textbf{ECAL-Driven Seeds} that have a matching \textbf{Sim-Track}
  92. \item \textbf{GSF Tracking Purity \--} The fraction of \textbf{GSF Tracks} that have a matching \textbf{Sim-Track}
  93. \item \textbf{GSF Tracking Fake Rate \--} The fraction of nontruth-matched Super-Clusters which result in at least one \textbf{GSF Track}.
  94. \end{itemize}
  95. \end{frame}
  96. % \begin{frame}{Previous status-quo}
  97. % \begin{columns}
  98. % \begin{column}{0.45\textwidth}
  99. % {\small
  100. % \begin{itemize}
  101. % \item In a previous presentation\footnotemark, I showed efficiency vs. purity for
  102. % \begin{itemize}
  103. % \item Old pair-match seeding (\texttt{ElectronSeedProducer})
  104. % \item New triplet seeding (\texttt{ElectronNHitSeedProducer}) for several choices of matching windows.
  105. % \end{itemize}
  106. % \item Performance of new seeding at the \texttt{wide} working point was comparable to old seeding in low-fake ($Z\rightarrow e^+e^-$) environment
  107. % \item Needed to validate performance in a high fake environment.
  108. % \end{itemize}
  109. % }
  110. % \end{column}
  111. % \begin{column}{0.6\textwidth}
  112. % \begin{figure}
  113. % \includegraphics[width=0.9\textwidth]{../common/figures/tracking_roc_curves_linear_plus_old_hoe.png}
  114. % \end{figure}
  115. % \end{column}
  116. % \end{columns}
  117. % \footnotetext[1]{\tiny \url{https://indico.cern.ch/event/697077/contributions/2936039/attachments/1618649/2573874/main.pdf}}
  118. % \end{frame}
  119. \begin{frame}{Relative Performance \-- GSF Tracking Efficiency}
  120. \begin{columns}
  121. \begin{column}{0.5\textwidth}
  122. \begin{itemize}
  123. \item Figure shows GSF Tracking efficiency vs kinematic variables of the electron \texttt{SimTrack}
  124. \item Efficiency is comparable for both DY and $t\bar{t}$ environments and for both algorithms and working points.
  125. \item Largest differences appear at low $p_T$ and in the barrel/endcap transition region.
  126. \end{itemize}
  127. \end{column}
  128. \begin{column}{0.5\textwidth}
  129. \begin{figure}
  130. GSF Tracking Efficiency
  131. \includegraphics[width=1.0\textwidth]{live_figures/tracking_eff_dR.png}
  132. \end{figure}
  133. \end{column}
  134. \end{columns}
  135. \blfootnote{\tiny This and the following slide have been show before and are included for completeness}
  136. \end{frame}
  137. \begin{frame}{Relative Performance \-- GSF Track Purity}
  138. \begin{columns}
  139. \begin{column}{0.5\textwidth}
  140. \begin{itemize}
  141. \item Figure shows GSF Tracking purity vs kinematic variables of the \texttt{GSFTrack}
  142. \item Clearly purity is affected by the higher fake environment in the $t\bar{t}$ sample.
  143. \item Note how the \texttt{default} working point of the new seeding (red/pink) has significantly better purity than the \texttt{} working point or the old seeding.
  144. \end{itemize}
  145. \end{column}
  146. \begin{column}{0.5\textwidth}
  147. \begin{figure}
  148. GSF Tracking Purity
  149. \includegraphics[width=1.0\textwidth]{live_figures/tracking_pur_dR.png}
  150. \end{figure}
  151. \end{column}
  152. \end{columns}
  153. \end{frame}
  154. \begin{frame}{Relative Performance \-- GSF Tracking Fake Rate}
  155. \begin{columns}
  156. \begin{column}{0.5\textwidth}
  157. \begin{itemize}
  158. \item Figure shows GSF Tracking fake rate vs kinematic variables of the supercluster
  159. \item Supercluster must have $HOE<0.15$, so fake are presumably from mostly photons or $\pi^0$
  160. \item There is a clear reduction in the fake rate with respect to the old method in both the \texttt{default} and \texttt{wide} working points.
  161. \item Seen in both $Z\rightarrow ee$ and $t\bar{t}$
  162. \end{itemize}
  163. \end{column}
  164. \begin{column}{0.5\textwidth}
  165. \begin{figure}
  166. GSF Tracking Fake Rate
  167. \includegraphics[width=1.0\textwidth]{live_figures/fake_rate_no_e_match.png}
  168. \end{figure}
  169. \end{column}
  170. \end{columns}
  171. \end{frame}
  172. \begin{frame}{Relative Performance \-- Prompt Efficiency}
  173. \begin{columns}
  174. \begin{column}{0.4\textwidth}
  175. \begin{itemize}
  176. \item The fraction of prompt electrons that match a GSF-Track
  177. \item Biggest improvements, again, happen at low $p_T$ and in the barrel/endcap transition region
  178. \item Note the change in the first bin relative to the overall efficiency (Slide 5). Large non-prompt contribution at low $p_T$.
  179. \end{itemize}
  180. \end{column}
  181. \begin{column}{0.6\textwidth}
  182. \begin{figure}
  183. Prompt GSF Tracking Efficiency
  184. \includegraphics[width=1.0\textwidth]{live_figures/prompt_eff_dR.png}
  185. \end{figure}
  186. \end{column}
  187. \end{columns}
  188. \end{frame}
  189. \begin{frame}{Relative Performance \-- Seed Multiplicity}
  190. \begin{columns}
  191. \begin{column}{0.4\textwidth}
  192. \begin{itemize}
  193. \item A single supercluster can potentially produce many seeds if it matches with many nearby tracks, however only one of these can be from the electron.
  194. \item Reducing the number of overall seeds while still producing \emph{the} correct one is desirable from a computational perspective.
  195. \item The new seeding scheme (\texttt{wide} WP) reduces the number of seeds by a factor of 3.8 for $t\bar{t}$ and 5.6 for $Z\rightarrow ee$.
  196. \end{itemize}
  197. \end{column}
  198. \begin{column}{0.6\textwidth}
  199. \begin{figure}
  200. Number of Electron Seeds Per Event
  201. \includegraphics[width=1.0\textwidth]{live_figures/number_of_good_seeds.png}
  202. \end{figure}
  203. \end{column}
  204. \end{columns}
  205. \end{frame}
  206. \begin{frame}{Overall Performance}
  207. \begin{center}
  208. Integrating over all tracks with $p_T>20$GeV and $|\eta|<2.5$ yields the performance numbers below.
  209. \begin{figure}
  210. % Number of Electron Seeds Per Event
  211. \includegraphics[width=0.6\textwidth]{figures/eff_table.png}
  212. \end{figure}
  213. \begin{itemize}
  214. \item The HLT default settings (\texttt{new-default}) of the new pixel matching
  215. scheme yield non-trivially better purity at the loss of some efficiency
  216. with respect to both the old seeding and the \texttt{wide} working point.
  217. \item The \texttt{wide} working point of the new seeding matches the
  218. \texttt{old-seeding} within errors except for purity is $\approx 2$\%
  219. better in the $t\bar{t}$ sample
  220. \item Most likely better to choose the \texttt{wide} working point over the \texttt{default} one to get the gain in efficiency, and count on subsequent filters to compensate for the worse purity and fake rate.
  221. \end{itemize}
  222. \end{center}
  223. \end{frame}
  224. % ask for conclusion to project and find out
  225. % - what changes need to be made
  226. % - who is going to implement them
  227. \begin{frame}{Conclusions \& Outlook}
  228. \begin{itemize}
  229. \item The new seeding algorithm has been optimized to have better or comparable performance to the current Offline seeding method(\texttt{old-default}) in all investigated metrics including
  230. \begin{itemize}
  231. \item GSF Tracking Efficiency
  232. \item GSF Tracking Purity
  233. \item GSF Tracking Fake Rate
  234. \item Number of Seeds
  235. \end{itemize}
  236. \item Unless there are objections, propose to move forward with implementing the new algorithm as the default in the next available CMSSW release.
  237. \end{itemize}
  238. \blfootnote{\tiny Analysis and ploting code is available at \url{https://git.fangmeier.tech/caleb/EGamma\_ElectronTrackingValidation}}
  239. \blfootnote{\tiny Additional plots are available at \url{https://eg.fangmeier.tech/seeding\_studies\_2018\_06\_20\_17/hists.html}}
  240. \end{frame}
  241. \appendix
  242. \backupbegin%
  243. \begin{frame}
  244. \begin{center}
  245. {\Huge BACKUP}
  246. \end{center}
  247. \end{frame}
  248. \begin{frame}{Overall Performance}
  249. \begin{columns}
  250. \begin{column}{0.5\textwidth}
  251. \begin{figure}
  252. GSF Tracking Performance (Hit Matched)
  253. \includegraphics[width=1.0\textwidth]{live_figures/tracking_roc_curve.png}
  254. \end{figure}
  255. \end{column}
  256. \begin{column}{0.5\textwidth}
  257. \begin{figure}
  258. GSF Tracking Performance ($\Delta R$ Matched)
  259. \includegraphics[width=1.0\textwidth]{live_figures/tracking_roc_curve_dR.png}
  260. \end{figure}
  261. \end{column}
  262. \end{columns}
  263. \end{frame}
  264. \begin{frame}{Matching Window Parameters}
  265. \begin{table}[]
  266. \centering
  267. \begin{tabular}{@{}llrrrr@{}}
  268. \toprule
  269. & & \textbf{narrow} & \textbf{default (HLT)} & \textbf{wide} & \textbf{extra-wide} \\ \midrule
  270. Hit 1 & dPhiMaxHighEt & \textbf{0.025} & \textbf{0.05} & \textbf{0.1} & \textbf{0.15} \\
  271. & dPhiMaxHighEtThres & 20.0 & 20.0 & 20.0 & 20.0 \\
  272. & dPhiMaxLowEtGrad & -0.002 & -0.002 & -0.002 & -0.002 \\
  273. & dRzMaxHighEt & 9999.0 & 9999.0 & 9999.0 & 9999.0 \\
  274. & dRzMaxHighEtThres & 0.0 & 0.0 & 0.0 & 0.0 \\
  275. & dRzMaxLowEtGrad & 0.0 & 0.0 & 0.0 & 0.0 \\ \midrule
  276. Hit 2 & dPhiMaxHighEt & \textbf{0.0015} & \textbf{0.003} & \textbf{0.006} & \textbf{0.009} \\
  277. & dPhiMaxHighEtThres & 0.0 & 0.0 & 0.0 & 0.0 \\
  278. & dPhiMaxLowEtGrad & 0.0 & 0.0 & 0.0 & 0.0 \\
  279. & dRzMaxHighEt & \textbf{0.025} & \textbf{0.05} & \textbf{0.1} & \textbf{0.15} \\
  280. & dRzMaxHighEtThres & 30.0 & 30.0 & 30.0 & 30.0 \\
  281. & dRzMaxLowEtGrad & -0.002 & -0.002 & -0.002 & -0.002 \\ \midrule
  282. Hit 3+ & dPhiMaxHighEt & \textbf{0.0015} & \textbf{0.003} & \textbf{0.006} & \textbf{0.009} \\
  283. & dPhiMaxHighEtThres & 0.0 & 0.0 & 0.0 & 0.0 \\
  284. & dPhiMaxLowEtGrad & 0.0 & 0.0 & 0.0 & 0.0 \\
  285. & dRzMaxHighEt & \textbf{0.025} & \textbf{0.05} & \textbf{0.1} & \textbf{0.15} \\
  286. & dRzMaxHighEtThres & 30.0 & 30.0 & 30.0 & 30.0 \\
  287. & dRzMaxLowEtGrad & -0.002 & -0.002 & -0.002 & -0.002 \\ \bottomrule
  288. \end{tabular}
  289. \end{table}
  290. \centering
  291. \texttt{NHit} Seeding window parameters. Bold designates modified values.
  292. \end{frame}
  293. \begin{frame}{Samples}
  294. \begin{itemize}
  295. \item {\tiny /ZToEE\_NNPDF30\_13TeV-powheg\_M\_120\_200/RunIISummer17DRStdmix-NZSFlatPU28to62\_92X\_upgrade2017\_realistic\_v10-v1}
  296. \item {\tiny /TT\_TuneCUETP8M2T4\_13TeV-powheg-pythia8/RunIISummer17DRStdmix-NZSFlatPU28to62\_92X\_upgrade2017\_realistic\_v10-v2}
  297. \end{itemize}
  298. \end{frame}
  299. \backupend%
  300. \end{document}