vallis / kissreport Goto Github PK

We might want to mention that this is an example of "blocked Gibbs" sampling. I can provide a higher quality version of the image

In the discussions (with Anne, Travis, Curt and others) we talked about parallel architectures for stochastic computing. One theme was that parallel tempering can be completely asynchronous, so with thousands of chains running in parallel there should always be pairs that can propose swaps. Another theme was mixed CPU/GPU architectures with GPUs being able to do e.g. wavelet or FFTs for likelihood calculation.

I think that Jeff Crowder's thesis work on applying MCMCs to the galactic binary problem should be referenced. It is still the best we have been able to do for the LISA global analysis problem (the new code that Tyson, Travis and I have put together does outperform Jeff's, but isn't published yet).

@Article{Crowder:2006eu,
author = "Crowder, Jeff and Cornish, Neil",
title = "{A Solution to the Galactic Foreground Problem for LISA}",
journal = "Phys. Rev.",
volume = "D75",
year = "2007",
pages = "043008",
doi = "10.1103/PhysRevD.75.043008",
eprint = "astro-ph/0611546",
archivePrefix = "arXiv",
primaryClass = "astro-ph",
SLACcitation = "%%CITATION = ASTRO-PH/0611546;%%"
}

For the waveform accuracy I suggest adding the comment that for the systematic error to be below the statistical error we need the mis-match to be less than D/(2 SNR^2), where D is the number of parameters in the waveform. The reference is

@Article{Chatziioannou:2016ezg,
author = "Chatziioannou, Katerina and Klein, Antoine and Cornish,
Neil and Yunes, Nicolas",
title = "{Analytic Gravitational Waveforms for Generic Precessing
Binary Inspirals}",
journal = "Phys. Rev. Lett.",
volume = "118",
year = "2017",
number = "5",
pages = "051101",
doi = "10.1103/PhysRevLett.118.051101",
eprint = "1606.03117",
archivePrefix = "arXiv",
primaryClass = "gr-qc",
SLACcitation = "%%CITATION = ARXIV:1606.03117;%%"
}

There is a new paper on this topic that I think we should reference. We showed that the difference in phasing for glitches and signals in the TDI variables will allow us to distinguish unmodeled gravitational wave burst from instrument glitches. It is just a first stab at the problem, and a lot more work has to be done, but it is at least a start.

@Article{Robson:2018jly,
author = "Robson, Travis and Cornish, Neil J.",
title = "{Detecting Gravitational Wave Bursts with LISA in the
presence of Instrumental Glitches}",
journal = "Phys. Rev.",
volume = "D99",
year = "2019",
number = "2",
pages = "024019",
doi = "10.1103/PhysRevD.99.024019",
eprint = "1811.04490",
archivePrefix = "arXiv",
primaryClass = "gr-qc",
SLACcitation = "%%CITATION = ARXIV:1811.04490;%%"
}