MEETING-DATE:September 30, 2010 MEETING-LOCATION:DC 1331 MEETING-TIME:10:30 MEETING-CHAIR:Philippe Lamoureux MEETING-CHAIRPIC:philippe.jpg COFFEE-HOUR-LAST-WEEK:Tiffany COFFEE-HOUR-THIS-WEEK:NOT Marshall COFFEE-HOUR-NEXT-WEEK:Volunteers? FORTH-DATE1:October 7, 2010 FORTH-DATE2:October 14, 2010 FORTH-DATE3:October 21, 2010 FORTH-DATE4:October 28, 2010 FORTH-LOCATION1:DC 1331 10:30 FORTH-LOCATION2:NOT DC 1331 10:30 FORTH-LOCATION3:DC 1331 1:30 FORTH-LOCATION4:DC 1331 10:30 FORTH-CHAIR1:Ben Lafreniere FORTH-CHAIR2:Tiffany Inglis FORTH-CHAIR3:Ed Lank FORTH-CHAIR4:Stephen Mann FORTH-CHAIRPIC1:ben.jpg FORTH-CHAIRPIC2:tiffany.jpg FORTH-CHAIRPIC3:ed.jpg FORTH-CHAIRPIC4:mann.jpg FORTH-TP1:Zainab AlMeraj FORTH-TP2:Tiffany Inglis FORTH-TP3:Ben Lafreniere FORTH-TP4:Philippe Lamoureux FORTH-TPPIC1:zainab.jpg FORTH-TPPIC2:tiffany.jpg FORTH-TPPIC3:ben.jpg FORTH-TPPIC4:philippe.jpg TPNAME:Gabriel Esteves TPTITLE:Run-time generation of QMC-Kalman filters for track fitting TPABSTRACT:One of the bottlenecks of the pattern recognition task in High Energy Physics is that of on-line track reconstruction. This has been traditionally divided into the sub-tasks of track finding and track fitting. The latter involves estimating the state of a particle inside a detector moving under the influence of a magnetic field. For the last twenty or so years, the most popular solution to the track fitting problem is the Kalman filter (KF). As powerful as it is, the assumptions under which the KF is guaranteed to compute the optimal estimator are not met in the track fitting problem. In particular, the dynamics are clearly non-linear and the process and measurement noise in the model are strongly non-Gaussian due to effects such as multiple Coulomb scattering and energy loss. A proposed solution is the "Gaussian sum" filter (GSF), which runs a bank of KFs to estimate each of the modes of the noise distribution, modeled here as a Gaussian mixture. In this paper, we take advantage of Intel's recent parallel frameworks dynamic code generation features to create a GSF that matches the given (observation) noise distribution. We further combat non-linearity by having the GSF drive, instead of KFs, the recently proposed quasi-Monte Carlo Kalman Filters, a generalization of the sigma-point KFs. The generated code is not only tailored to the data, but takes advantage of several levels of parallelism in multi-core processors. TPPIC:rgesteve.jpg DIONE: DITWO: DITHREE: DIFOUR: AIONE:Poster! AITWO:DC 1331 is not available on October 14, but DC 1316 is. DC 1331 is not available on December 2, and neither is anything else (except perhaps the lab). AITHREE:Graphically enhance the 488 course? AIFOUR:What language should replace the hideous "language" Scheme for 1st year courses? DMONE: DMTWO: DMTHREE: DMFOUR: SEMINARS: