GPU-accelerated Parameter Estimation

Document #:

JGW-G1807674-v2

Document type:

G

Submitted by:

Sadakazu Haino

Updated by:

Sadakazu Haino

Document Created:

19 Jan 2018, 13:26

Contents Revised:

19 Jan 2018, 13:53

DB Info Revised:

16 Mar 2018, 05:46

Abstract:

Parameter Estimation (PE) is a time-consuming process; we need to find a new way to accelerate it in view of many-GW-detection era towards 3G. A Frequency-domain CBC PE on 15 parameters with Nested sampling with MCMC sub-chains and IMRPhenomPv2 waveform model has been accelerated with GPU by factor of ~100. The codes are all newly written in C++ and CUDA and produce the same output as LALInference for cbcBayesPostProc

Files in Document:

• Slides file (gpe.pdf, 1.1 MB)

Topics:

• Activity:KAGRA
• Management:Collaboration meeting:KAGRA meeting
• Detector:Data Analysis

Authors:

• Sadakazu Haino

Referenced by:

• JGW-T1808090: O3 Sky Localization Calculation Comparison between MCMC and Fisher Analysis