I'm curious as to what exactly my computer is processing when it completes Gravitational wave searches or Gamma ray pulse search work units?
There is some material online that will explain this. May I ask how familiar you are with physics in general so we can point you to the right kind of description?
I'm actually an Engineering student, I have a sound understanding of physics.
Aha! :-)
In that case we are convolving signal templates with a given data stream, per work unit. We receive time series signals from the respective detectors, whip them through an FFT and slap that against a bank of candidates of putative signal shapes. Because we will almost never predict the exact template that will fit some as-yet-undiscovered signal, what we generate is a statistic that reflects the merit of the degree of fit b/w data stream and template. Thus a given work unit returns to our servers a list of highest match candidates for further examination. The spacing of the templates in the relevant parameter space is such as to reasonably minimise the odds on missing real signals, of course balancing that against false positives and yet remain in overall the bounds of an achievable project workload.
From a total project workflow point of view, the work units processed by our volunteers represent the 'embarrassingly parallel' part. Meaning that one can readily produce template banks, signal FFT's, and perform candidate list triage ( and promising follow-ups ! ) pretty much all on the server side. But the work units effectively do the convolution and match-statistic generation portions.
The detail is, as ever, rather more messier than this precis. But that's the guts of it. I might add too that it actually works. We have found several dozen signals that have been confirmed to satisfactory precision by specifically targeted subsequent observations. If you like, E@H acts as tripwire for localising real sky sources sufficiently to be then exposed to far more accurate examinations.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
What exactly does my computer do S6 WU's
)
There is some material online that will explain this. May I ask how familiar you are with physics in general so we can point you to the right kind of description?
Cheers
HB
I'm actually an Engineering
)
I'm actually an Engineering student, I have a sound understanding of physics.
RE: I'm actually an
)
Aha! :-)
In that case we are convolving signal templates with a given data stream, per work unit. We receive time series signals from the respective detectors, whip them through an FFT and slap that against a bank of candidates of putative signal shapes. Because we will almost never predict the exact template that will fit some as-yet-undiscovered signal, what we generate is a statistic that reflects the merit of the degree of fit b/w data stream and template. Thus a given work unit returns to our servers a list of highest match candidates for further examination. The spacing of the templates in the relevant parameter space is such as to reasonably minimise the odds on missing real signals, of course balancing that against false positives and yet remain in overall the bounds of an achievable project workload.
From a total project workflow point of view, the work units processed by our volunteers represent the 'embarrassingly parallel' part. Meaning that one can readily produce template banks, signal FFT's, and perform candidate list triage ( and promising follow-ups ! ) pretty much all on the server side. But the work units effectively do the convolution and match-statistic generation portions.
The detail is, as ever, rather more messier than this precis. But that's the guts of it. I might add too that it actually works. We have found several dozen signals that have been confirmed to satisfactory precision by specifically targeted subsequent observations. If you like, E@H acts as tripwire for localising real sky sources sufficiently to be then exposed to far more accurate examinations.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Thanks mike, Well that
)
Thanks mike,
Well that makes sense. I had a gut feeling that it was doing that.