https://setiathome.berkeley.edu/forum_thread.php?id=82673#1923224

If we detect multiplets of mixed types (spike+gaussian or pulse+triplet) these will have better scores than the corresponding singe-type multiplets. They will be more likely to stand out from noise multiplets, and we'll have a better chance of finding such an ET signal.

Implementing this change was a major effort, since all the code was designed to work with one signal type at a time. I removed this restriction, and took the opportunity to do an overall cleanup on the code, including removing all database dependencies. I also changed things so that we can use the same executables for both S@h and SERENDIP; previously we had to compile them separately.

So, after reading this thread I'm assuming he's talking about how they analyze our returned results??

Not, the way we ourselves analyze the work units sent to us. Anyone??

Cause if it's our method of checking, then I have to wonder how much of the data already returned would have to be redone. Not to mention those of use that have installed Lunatics applications which are not sent to use by the Server but installed ourselves. Having been installed ourselves, whatever changes were made to Server sent applications to change analysis would not occur with Lunatics installed applications or privately built applications (I'm really thinking it's screening our results)

Either way, I haven't seen anyone comment on this in that thread and thought maybe we could discuss it here and then if needed, send a inquire to DA if it really is the latter.

Z

---

The Nebula project is the next step in the process of identifying potential alien RF sources.
Our process is to screen out the most unlikely data and identify the (relatively) few "signals worth further processing". Our results are "simple" lists of signal types, centre frequency spread/split frequencies and other "simple" data. All of which should be very independent of the application employed by us PROVIDED that application has been correctly validated before deployment and that there is enough randomness in the selection of that application(*).
Nebula takes this data and does some correction to it to "normalise" it, that is remove frequency errors due to the motion of the telescope, Earth's orbit on its own axis, Earth's orbit about the Sun. This step requires a fair bit of high precision arithmetic to make appear that the Earth and telescope were stationary in space.
Next Nebula compares sets of data from the same location in space, using the techniques David Anderson outlines in the post you linked to. This entails more corrections for drift (real and potential).
It is possible that, in the future "near miss" candidate work will be reprocessed, but I doubt that will be a significant effort, or that there will be very much of it.


(*) This is why "cherry-picking" and CPU/GPU rescheduling are not "good ideas" as the both reduce the randomness.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Single-type multiplets.