|
Participant |
Pinto, Innocenzo M | |||||||
|
Institution |
University of Sannio, INFN, LVC and KAGRA - C.so Garibaldi 107 - Benevento - - - Italy | |||||||
|
Session |
PT3 |
Accepted |
Yes |
Order |
99 |
Time |
||
|
Talk |
Poster abstract |
Title |
Glitch Entomology | |||||
| Coauthors | Fusco, Adele; Mejuto-Villa, Elena, Pinto, Innocenzo M.; Principe, Maria; Troiano, Luigi | |||||||
|
Abstract |
Glitches are transient disturbances of environmental/instrumental origin observed ubiquitously in the data of interferometric detectors of gravitational waves. Their occurrence entails substantial deviations from the usual stationary-Gaussian noise floor assumption, making the noise background non-stationary and non-Gaussian. Principal component analysis shows that glitches live in a low-dimensional functional space, and the natural question is posed of finding the best "dictionary" to represent them. We propose a physically motivared multiple-input/multiple-output (MIMO) model of an interferometric detector, whereby glitches in any output channel are superpositions of a small number of canonical responses, specific to that channel, with random amplitudes and delays. Generalization to nonlinear input/output models is straightforward via Volterra series. We show that those canonical responses can be identified from a sufficiently large set of glitchy data using independent component analysis (ICA), opening the way to active, almost real time glitch subtraction. |
|||||||
Pdf file |
|
|||||||