The development of the radio pulsar search app is mainly Olivers responsibility, not mine. I myself won't put much time into that that distracts me from the GW search. I definitely don't want to rush things, I just wanted to keep you informed. Probably the time estimates will turn out to be too optimistic, as usual.
The development of the radio pulsar search app is mainly Olivers responsibility, not mine. I myself won't put much time into that that distracts me from the GW search. I definitely don't want to rush things, I just wanted to keep you informed. Probably the time estimates will turn out to be too optimistic, as usual.
BM
All info is very welcome. You and your colleagues are doing great job and you will make it right. Thanks for that.
The old ABP2 App has a flaw in the harmonic summing code that costs a bit of sensitivity (~2%).
Just to clarify (as questions were already asked), this is about the resampling stage, not the harmonic summing. It's also not a bug but a different, non-optimal physical approach being used. We improved the resampling algorithm such that these 2% in sensitivity are not lost anymore.
I realize that I could technically find this in the source myself, but I haven't taken the time to familiarize myself with the repository yet. What resampling algorithm were you using before, and what are you using now? Something general like cubic or hermite spline, or something more specialized?
I definitely don't want to rush things, I just wanted to keep you informed.
I just want to say how much I appreciate the great work you all do. It is nice to have such a great level of professionalism in this project. No need to rush things before they are ready!
I realize that I could technically find this in the source myself, but I haven't taken the time to familiarize myself with the repository yet. What resampling algorithm were you using before, and what are you using now? Something general like cubic or hermite spline, or something more specialized?
Hi,
we're doing a simple nearest neighbour resampling, which turns out to be very good for our purposes. In simulations we could show that only ~1% of signal power is lost compared to a perfect analysis. With the latest changes in the resampling code, we're now doing the nearest neighbour resampling optimally. Before, we used an approximation in the signal model that would lose ~3% of signal power. So, no really big change.
With the latest changes in the resampling code, we're now doing the nearest neighbour resampling optimally.
I see ... I look forward to looking at the code at some point :) I am a little surprised that even a simple linear spline isn't worth the speed hit - something like "y[i] + (x - x[i])*(y[i+1] - y[i])/(x[i+1] - x[i])"?
Will ABP2 ever be finished?
5 Beams/day is not much...
or will the remaining tasks be send out again, if they are not computed by the ones that got them?
Right now there are only 1,408 tasks in progress. One beam has 50 to 60 tasks. So there are only about 25 beams left. So only one week to go. We won´t reach 100%. But this could be due to some problems Bernd mentioned
Quote:
In addition there are a few data sets which we couldn't process with our current workunit generator, e.g. because teh data was too noisy. We'll try to feed this data into the new workunit generator when it's ready, but I doubt that we'll find something in these datasets anyway.
With the latest changes in the resampling code, we're now doing the nearest neighbour resampling optimally.
I see ... I look forward to looking at the code at some point :) I am a little surprised that even a simple linear spline isn't worth the speed hit - something like "y[i] + (x - x[i])*(y[i+1] - y[i])/(x[i+1] - x[i])"?
That would be a low-passs filter in frequency which is something we don't want to have; it limits our ability to see the high frequency components.
The development of the radio
)
The development of the radio pulsar search app is mainly Olivers responsibility, not mine. I myself won't put much time into that that distracts me from the GW search. I definitely don't want to rush things, I just wanted to keep you informed. Probably the time estimates will turn out to be too optimistic, as usual.
BM
BM
RE: The development of the
)
All info is very welcome. You and your colleagues are doing great job and you will make it right. Thanks for that.
RE: The old ABP2 App has a
)
Just to clarify (as questions were already asked), this is about the resampling stage, not the harmonic summing. It's also not a bug but a different, non-optimal physical approach being used. We improved the resampling algorithm such that these 2% in sensitivity are not lost anymore.
Cheers,
Oliver
Einstein@Home Project
I realize that I could
)
I realize that I could technically find this in the source myself, but I haven't taken the time to familiarize myself with the repository yet. What resampling algorithm were you using before, and what are you using now? Something general like cubic or hermite spline, or something more specialized?
RE: I definitely don't want
)
I just want to say how much I appreciate the great work you all do. It is nice to have such a great level of professionalism in this project. No need to rush things before they are ready!
RE: I realize that I could
)
Hi,
we're doing a simple nearest neighbour resampling, which turns out to be very good for our purposes. In simulations we could show that only ~1% of signal power is lost compared to a perfect analysis. With the latest changes in the resampling code, we're now doing the nearest neighbour resampling optimally. Before, we used an approximation in the signal model that would lose ~3% of signal power. So, no really big change.
Cheers, Ben
Einstein@Home Project
RE: With the latest changes
)
I see ... I look forward to looking at the code at some point :) I am a little surprised that even a simple linear spline isn't worth the speed hit - something like "y[i] + (x - x[i])*(y[i+1] - y[i])/(x[i+1] - x[i])"?
Will ABP2 ever be finished? 5
)
Will ABP2 ever be finished?
5 Beams/day is not much...
or will the remaining tasks be send out again, if they are not computed by the ones that got them?
Right now there are only
)
Right now there are only 1,408 tasks in progress. One beam has 50 to 60 tasks. So there are only about 25 beams left. So only one week to go. We won´t reach 100%. But this could be due to some problems Bernd mentioned
RE: RE: With the latest
)
That would be a low-passs filter in frequency which is something we don't want to have; it limits our ability to see the high frequency components.
Einstein@Home Project