[Forum SIS] Annuncio di Seminario
Claudio Durastanti
claudio.durastanti a unipv.it
Lun 22 Ott 2012 11:30:49 CEST
Ciclo di Seminari - Progetto ERC PASCAL (Probabilistic and Statistical
Techniques for Cosmological Applications)
Lunedì 29 Ottobre ore 16
Aula Dal Passo
Dipartimento di Matematica
Università di Roma Tor Vergata
Speaker:
-----------
Armin Schwartzman
Harvard School of Public Health
Title:
----------
Multiple Testing of Local Maxima for Detection of Peaks in 1D
Abstract:
----------
A common problem in signal/image analysis is to find local significant
regions, either for a single signal/image or for the difference
between two or more signals/images. In this talk, I describe how to
approach this inference problem from a multiple testing point of view,
and emphasize the need to make inferences about spatial features such
as peaks rather than individual pixels or voxels. Focusing on the 1D
case, I propose a formal procedure for detecting smooth peaks buried
in stationary noise, where both the height and location of the peaks
are unknown. The procedure, involving kernel smoothing and testing of
local maxima, is easy to implement and takes advantage of existing
multiple testing procedures, so that global error rates are
controlled. Interestingly, the optimal bandwidth corresponds to the
"matched filter" principle, where the kernel size should be close to
that of the peaks to be detected. The method is illustrated in 1D time
series data of neuronal recordings.
--
------------------------------
Questa informativa è inserita in automatico dal sistema al fine esclusivo
della realizzazione dei fini istituzionali dell'ente.
Diventa anche tu sponsor dei nostri ricercatori. Scegli di destinare il 5
per mille all’Università di Pavia <http://www.unipv.it/>: offrirai nuove
opportunità alla ricerca, ai giovani e al territorio. Un gesto che non
costa nulla e costruisce tanto. C.F. dell’Università di Pavia 80007270186.
Please note that the above message is addressed only to individuals filing
Italian income tax returns.
Maggiori informazioni sulla lista
Sis