Comparing different searches for gravitational-wave bursts on simulated LIGO and VIRGO data Michele Zanolin -MIT on behalf of the LIGO-VIRGO joint working.

Slides:



Advertisements
Similar presentations
GWDAW 11 - Potsdam, 19/12/ Coincidence analysis between periodic source candidates in C6 and C7 Virgo data C.Palomba (INFN Roma) for the Virgo Collaboration.
Advertisements

GWDAW-8 (December 17-20, 2003, Milwaukee, Wisconsin, USA) Search for burst gravitational waves with TAMA data Masaki Ando Department of Physics, University.
Burst detection efficiency  In order to interpret our observed detection rate (upper limit) we need to know our efficiency for detection by the IFO and.
LIGO-G Z Coherent Coincident Analysis of LIGO Burst Candidates Laura Cadonati Massachusetts Institute of Technology LIGO Scientific Collaboration.
Systematic effects in gravitational-wave data analysis
G Z April 2007 APS Meeting - DAP GGR Gravitational Wave AstronomyKeith Thorne Coincidence-based LIGO GW Burst Searches and Astrophysical Interpretation.
LIGO-G Z Detector characterization for LIGO burst searches Shourov K. Chatterji for the LIGO Scientific Collaboration 10 th Gravitational Wave.
1/25 Current results and future scenarios for gravitational wave’s stochastic background G. Cella – INFN sez. Pisa.
1 Characterising the Space of Gravitational-Wave Bursts Patrick J. Sutton California Institute of Technology.
Observing the Bursting Universe with LIGO: Status and Prospects Erik Katsavounidis LSC Burst Working Group 8 th GWDAW - UWM Dec 17-20, 2003.
S.Klimenko, G Z, December 21, 2006, GWDAW11 Coherent detection and reconstruction of burst events in S5 data S.Klimenko, University of Florida.
LIGO-G Z LIGO Scientific Collaboration 1 Upper Limits on the Rate of Gravitational Wave Bursts from the First LIGO Science Run Edward Daw Louisiana.
A coherent null stream consistency test for gravitational wave bursts Antony Searle (ANU) in collaboration with Shourov Chatterji, Albert Lazzarini, Leo.
LIGO-G Z Peter Shawhan, for the LIGO Scientific Collaboration APS Meeting April 25, 2006 Search for Gravitational Wave Bursts in Data from the.
LIGO-G Z The AURIGA-LIGO Joint Burst Search L. Cadonati, G. Prodi, L. Baggio, S. Heng, W. Johnson, A. Mion, S. Poggi, A. Ortolan, F. Salemi, P.
LIGO-G M GWDAW, December LIGO Burst Search Analysis Laura Cadonati, Erik Katsavounidis LIGO-MIT.
M. Principe, GWDAW-10, 16th December 2005, Brownsville, Texas Modeling the Performance of Networks of Gravitational-Wave Detectors in Bursts Search Maria.
Multidimensional classification of burst triggers from the fifth science run of LIGO Soma Mukherjee CGWA, UTB GWDAW11, Potsdam, 12/18/06 LIGO-G
LIGO-G Z A Coherent Network Burst Analysis Patrick Sutton on behalf of Shourov Chatterji, Albert Lazzarini, Antony Searle, Leo Stein, Massimo.
The Analysis of Binary Inspiral Signals in LIGO Data Jun-Qi Guo Sept.25, 2007 Department of Physics and Astronomy The University of Mississippi LIGO Scientific.
S.Klimenko, December 2003, GWDAW Performance of the WaveBurst algorithm on LIGO S2 playground data S.Klimenko (UF), I.Yakushin (LLO), G.Mitselmakher (UF),
Abstract: We completed the tuning of the analysis procedures of the AURIGA-LIGO joint burst search and we are in the process of verifying our results.
Searching for Gravitational Waves with LIGO Andrés C. Rodríguez Louisiana State University on behalf of the LIGO Scientific Collaboration SACNAS
LIGO-G Z April 2006 APS meeting Igor Yakushin (LLO, Caltech) Search for Gravitational Wave Bursts in LIGO’s S5 run Igor Yakushin (LLO, Caltech)
S.Klimenko, July 14, 2007, Amaldi7,Sydney, G Z Detection and reconstruction of burst signals with networks of gravitational wave detectors S.Klimenko,
Amaldi-7 meeting, Sydney, Australia, July 8-14, 2007 LIGO-G Z All-Sky Search for Gravitational Wave Bursts during the fifth LSC Science Run Igor.
LIGO- G D Burst Search Report Stan Whitcomb LIGO Caltech LSC Meeting LIGO1 Plenary Session 18 August 2003 Hannover.
Searching for gravitational-wave bursts with the Q Pipeline Shourov K. Chatterji LIGO Science Seminar 2005 August 2.
G030XXX-00-Z Excess power trigger generator Patrick Brady and Saikat Ray-Majumder University of Wisconsin-Milwaukee LIGO Scientific Collaboration.
The Restricted Matched Filter for Distributed Detection Charles Sestok and Alan Oppenheim MIT DARPA SensIT PI Meeting Jan. 16, 2002.
HACR at Virgo: implementation and results Gabriele Vajente 12 th ILIAS WG1 meeting Geneva, March 29 th -30 th 2007.
S.Klimenko, August 2003, Hannover LIGO-G Z How optimal are wavelet TF methods? S.Klimenko l Introduction l Time-Frequency analysis l Comparison.
LIGO-G Data Analysis Techniques for LIGO Laura Cadonati, M.I.T. Trento, March 1-2, 2007.
LIGO-G Z The Q Pipeline search for gravitational-wave bursts with LIGO Shourov K. Chatterji for the LIGO Scientific Collaboration APS Meeting.
S.Klimenko, G Z, December 2006, GWDAW11 Coherent detection and reconstruction of burst events in S5 data S.Klimenko, University of Florida for.
S.Klimenko, G Z, December 21, 2006, GWDAW11 Coherent detection and reconstruction of burst events in S5 data S.Klimenko, University of Florida.
Joint LIGO-Virgo data analysis Inspiral and Burst Summary of the first project results Overview of the future activities M.-A. Bizouard (LAL-Orsay) on.
S.Klimenko, G Z, March 20, 2006, LSC meeting First results from the likelihood pipeline S.Klimenko (UF), I.Yakushin (LLO), A.Mercer (UF),G.Mitselmakher.
The 9th Gravitational Wave Data Analysis Workshop (December 15-18, 2004, Annecy, France) Results of the search for burst gravitational waves with the TAMA300.
LIGO-G Z Confidence Test for Waveform Consistency of LIGO Burst Candidate Events Laura Cadonati LIGO Laboratory Massachusetts Institute of Technology.
Data Analysis Algorithm for GRB triggered Burst Search Soumya D. Mohanty Center for Gravitational Wave Astronomy University of Texas at Brownsville On.
S.Klimenko, March 2003, LSC Burst Analysis in Wavelet Domain for multiple interferometers LIGO-G Z Sergey Klimenko University of Florida l Analysis.
Search for bursts with the Frequency Domain Adaptive Filter (FDAF ) Sabrina D’Antonio Roma II Tor Vergata Sergio Frasca, Pia Astone Roma 1 Outlines: FDAF.
A Comparison of Burst Gravitational Wave Detection Algorithms for LIGO Amber L. Stuver Center for Gravitational Wave Physics Penn State University.
S.Klimenko, December 2003, GWDAW Burst detection method in wavelet domain (WaveBurst) S.Klimenko, G.Mitselmakher University of Florida l Wavelets l Time-Frequency.
LIGO-G All-Sky Burst Search in the First Year of the LSC S5 Run Laura Cadonati, UMass Amherst For the LIGO Scientific Collaboration GWDAW Meeting,
Peter Shawhan The University of Maryland & The LIGO Scientific Collaboration Penn State CGWP Seminar March 27, 2007 LIGO-G Z Reaching for Gravitational.
Comparison of filters for burst detection M.-A. Bizouard on behalf of the LAL-Orsay group GWDAW 7 th IIAS-Kyoto 2002/12/19.
GWDAW91Thursday, December 16 First Comparison Between LIGO &Virgo Inspiral Search Pipelines F. Beauville on behalf of the LIGO-Virgo Joint Working Group.
Igor Yakushin, December 2004, GWDAW-9 LIGO-G Z Status of the untriggered burst search in S3 LIGO data Igor Yakushin (LIGO Livingston Observatory)
LIGO-G Z Searching for gravitational wave bursts with the new global detector network Shourov K. Chatterji INFN Sezioni di Roma / Caltech LIGO.
The first AURIGA-TAMA joint analysis proposal BAGGIO Lucio ICRR, University of Tokyo A Memorandum of Understanding between the AURIGA experiment and the.
G08XXXX-00-Z S. Chatterji, Cascina, Italy, 2008 November 261 Gravitational-wave data analysis and supernovae Shourov K. Chatterji INFN Sezione di Roma.
SEARCH FOR INSPIRALING BINARIES S. V. Dhurandhar IUCAA Pune, India.
Abstract: We completed the tuning of the analysis procedures of the AURIGA-LIGO joint burst search and we are in the process of verifying our results.
LIGO-G Z The Q Pipeline search for gravitational-wave bursts with LIGO Shourov K. Chatterji for the LIGO Scientific Collaboration APS Meeting.
LIGO-G05????-00-Z Detector characterization for LIGO burst searches Shourov K. Chatterji For the LIGO Scientific Collaboration 10 th Gravitational Wave.
The Q Pipeline search for gravitational-wave bursts with LIGO
Igor Yakushin, LIGO Livingston Observatory
Searching for gravitational wave bursts with the new global detector network 2007 May 3 Searching for gravitational wave bursts with the new global detector.
LIGO Scientific Collaboration meeting
M.-A. Bizouard, F. Cavalier
Maria Principe University of Sannio, Benevento, Italy
Targeted Searches using Q Pipeline
SLOPE: A MATLAB Revival
WaveBurst upgrade for S3 analysis
Coherent Coincident Analysis of LIGO Burst Candidates
A Waveform Consistency Test for Binary Inspirals using LIGO data
Status and Plans for the LIGO-TAMA Joint Data Analysis
Performance of the WaveBurst algorithm on LIGO S2 playground data
Presentation transcript:

Comparing different searches for gravitational-wave bursts on simulated LIGO and VIRGO data Michele Zanolin -MIT on behalf of the LIGO-VIRGO joint working group

Benefits Pipeline Results Future plans

Benefits of comparing LIGO and Virgo searches on single interferometer simulated data Compare time domain and frequency domain methods over different class of waveforms. Detector independent measure of performance Simulated noise (see inspiral talk) allows to test the limit of design sensitivity and improved vetos (and no restrictions due to use of real data) Single interferometer analysis decouples the properties of the trigger generation from those of coincidence modules Events might happen when only one interferometer is on lock

Pipeline 7 methods already involved (3 more will join soon) Simulated data have been produced Simulated injections have been generated Receiver operating characteristics have been computed on the 3 hours of simulated data for false rates from 0.1 to Hz. Trigger lists ( containing rough time-frequency volume) and sets of figures of merit have been generated A method independent parameterisation of the triggers is adopted (Sylvestre, Sutton, Lazzarini). A post processing stand alone parameter estimation module is introduced, that uses Maximum likelihood estimators and assumes colored Gaussian noise (Zanolin, Sylvestre).

Time-frequency Methods Q-pipeline (see talk by S.Chatterji): multiresolution time frequency search for excess power applied on data that are first whitened using zero phase linear prediction. Equivalent to optimal matched filter for minimum uncertainty waveforms of unknown phase in the whitened data. Kleine Welle (L.D. - L. Blackburn): search for statistically significant clusters of coefficients in the dyadic Haar wavelet decomposition. S-transform (V.D. - see talk byA.C. Clapson): search for statistically significant clusters of coefficients in the time frequency map generated using a kernel composed of complex exponentials shaped by Gaussian profiles with width inversely proportional to frequency. High pass and line removal applied on Virgo data. Power Filter (G.Guidi): search for excess power over different time intervals and sets of frequencies.

Time domain methods Peak Correlator (P.Hello): Search for peaks of Wiener filtered data with Gaussian templates. For Virgo data a high pass filter and a line removal filter are applied to remove the resonance at 0.6 Hz. Mean Filter (M. A. Bizouard): Search for excess in moving averages of whitened data over intervals containing from 10 to 200 samples. ALF (M.A.Bizouard): Search for change in slope over moving windows of whitened data over intervals containing from 10 to 300 samples.

Simulated Bursts Three hours sets of simulated bursts generated (peak times are Poisson distributed with mean rate of 1/60 seconds ) Bursts amplitude is parameterized in terms of h and snr (tested between 2 to10) Injected Gaussian have durations of 1 and 4 milliseconds Cosine Gaussian with central frequencies of Hz and Qs of 5 and 15 Supernovae core collapse waveforms (Dimmelmejer, Font, Muller) with parameters (a=1, b=2, g=1) and (a=2, b=4, g=1)

Burst events ● Three species: Dimmelmeier & al. a1b2g1,a2b4g1 Gaussian 1,4 milliseconds sine-gaussian shapes q=5,15 f=235,

Method independent event parametrization central frequency Bandwidth where duration where and h_rss. The stand alone parameter estimation gets the time frequency volumes from the searches and computes the burst’s parameters in the presence of noise

Injections locations in the parameter space cggadfm

Receiver operating characteristics for SG235Q5 snr=10 snr=5

Receiver operating characteristics for SG235Q5 snr=5 snr=10

Receiver operating characteristics for GA1d0 snr=10 snr=5

Receiver operating characteristics for GA1d0 snr=5

Receiver operating characteristics for dfm2b4g1 snr=10snr=5

Receiver operating characteristics for dfm2b4g1 snr=5

And/Or combination of methods l LIGO Single (snr=5) AND (snr=5) OR (snr=5) AND (snr=10) OR (snr=10) Single (snr=10) N injected sg235Q Q PF DFMa2b4g1 Virgo Q PC 84PC DFMa2b4g1GA1d PC 75PC Q Q/PC 170 MF/ALF/PC Q/ALF MF/ALF/PC 178 MF/ALF/PC/Q Each method is tuned at a false rate = 0.1 Hz Preliminary, since joint false rate still to be estimated. QQ

What did we learn Learn to process each others' data, exchange, and compare triggers Different methods perform differently in different regions of the parameter space OR combinations of methods increases efficiency (false rates for combinations of the methods still to be investigated) Confront burst search performance with optimal matched filtering (see S.Chatterji talk)

Future directions As first project of exchanging simulated noise and signal data of the LIGO and Virgo instruments is successfully carrying out and address the goals set by the joint working group the next steps are approaching: - A program to continue the data exchange within the framework of simulated data but using astrophysical (coherent) waveform injections onto simulations is forthcoming (useful to quantify the scientific potential of a combined analysis ) - Ultimate goal is to operate the instruments as part of a global network. We are defining the details of exchange of triggers (and/or) data will actually take place.

From now on back up material

Q-pipeline Targets miminum uncertaintly waveforms with 64<f [Hz]< <Q<50 Space of time, frequency, and Q is tilted for a worst case energy loss of 20 percent for minimum uncertainty waveforms The data are first whitened using zero-phase linear prediction Multi-resolution time-frequency search for excess power The whitened data is projected onto a basis of windowed complex exponentials where the window w(t) has mimimum time-frequency uncertainty and bandwidth Significant tiles are identified assuming white noise statistics Only the most significant set of non-overlapping tiles are reported Equivalent to an optimal matched filter search for minimum uncertainty waveforms of unknown phase in the whitened data stream See talk by S. Chatterji at this conference.

kleineWelle kleineWelle searches for statistically significant clusters of coefficients in the dyadic wavelet decomposition of a whitened timeseries.

Power Filter (PF) Compute the Power over several windows. Analyze the power in different sets of frequencies. Parameters of the analysis: Window 0: width = 0.03 s. Window 1: width =0.06 s, 10 sets of frequencies. Whitening of the data: Burg algorithm with a learning time of ~180 s 3000 paramaters for the LIGO simulated noise 1000 paramaters for the VIRGO simulated noise.

Peak Correlator (PC) PC : Wiener filtering with Gaussian templates Half-widths  are optimally computed in [0.2 ms; 6ms] according to a minimal match MM = 0.99 (  9 templates) In the case of the VIRGO simulated noise, a pre-processing is applied in order to kill the huge resonance at 0.6 Hz (Kalman filter).

Mean Filter (MF) MF computes the average of data in some moving window of size N : MF needs a whitening of the data. In the LIGOVIRGO MDC we have used a Burg algorithm with a learning time of 300 s and 3000 paramaters for the LIGO simulated noise and 4000 paramaters for the VIRGO simulated noise. The practical implementation has been done using 10 windows : N (bins) :

ALF ALF is based on a detection of change of slope in a window of size N. ALF needs a whitening of the data as MF (same algo and parameters). Assuming data being whitened, the slope a and offset b are computed : where means average over a window of size N The variables a and b are then decorrelated using variances and covariance being : All the information (slope and offset) is then combined in ALF : The practical implementation has been done using 10 windows : N (bins) :

Preprocessing (window length) –High pass filter + Hanning window. –Change to frequency domain (FFT). Build TF map using S-Transform (resolution) (for each frequency) –Product with Gaussian window. –Bring back to time domain (FFT -1 ). Post processing (for each frequency) –Normalize to zero mean and unit standard deviation. –Threshold map element value. (1 st threshold) Data reduction –Clustering of map elements (8 connectivity) –Threshold on total cluster energy. (2 nd threshold) S transform –Related to the Gabor (or wavelet) transform –Good signal localization (time and frequency) –Restitution of noise psd. –Invertible. S transform analysis chain Signal visible in map… Issue is reliable extraction!