"Known Lenses" not being classified?

[drphilmarshall]

We write in paper 2 section 7.3.1:

"Surprisingly, most of the lenses in the known sample have few classifications (Nclass<10) which includes both the detected and missed lenses."

But then we say:

"Overall, we do not find any correlation for the detected and the missed sample with Nclass."

I think these statements need clarifying in order to be consistent. Correlation between what and what?

Could people be going straight to Talk rather than marking the image, when they see a known lens? This could be tested with a future Talk comment text analysis.

[anupreeta27]

stage 1 report says mean classn/test subject is 8.4. so, i don't think KLs are special in any way. My "surprisingly ..few classfn" was probably either wrt to Nclass_max=50 or for the fact, that this was similar for both detected and missed lenses. i was expecting that probably missed lenses would have fewer Nclass than the detected ones but this was not the case.

[aprajita]

so will you edit the text in that section accordingly then? hope you manage to get some sleep before our telecon…

On 8 May 2015, at 14:34, anupreeta27 notifications@github.com wrote:

stage 1 report says mean classn/test subject is 8.4. so, i don't think KLs are special in any way. My "surprisingly ..few classfn" was probably either wrt to Nclass_max=50 or for the fact, that this was similar for both detected and missed lenses. i was expecting that probably missed lenses would have fewer Nclass than the detected ones but this was not the case.

— Reply to this email directly or view it on GitHub.

[ccld]

What was your take on known lenses overlapped by training sims?

[drphilmarshall]

It was good that we put those images into the system without sims as well! The false negative effect we are discussing here is not affected by the sim placement, I think.

On Sat, May 9, 2015 at 3:31 PM, ccld notifications@github.com wrote:

What was your take on known lenses overlapped by training sims?

— Reply to this email directly or view it on GitHub https://github.com/drphilmarshall/SpaceWarps/issues/202#issuecomment-100496944 .

[ccld]

I can't find images without sims of those overlapping known lenses

sim overlapping #known

SA11 ASW00015xe SA52 ASW0004aju SA68 ASW0000joc SA86 ASW0006842 SA89 ASW0006eoq SA115 ASW0007en2 SA127 ASW00093k7 SL2SJ140614+520252 ASW0005nx7

and for sim & known in same frame SL2S J084847-035103 ASW0004wuj

Could it yield an effect on known classification?

[anupreeta27]

@ccld not to worry, they are there. we are not missing any lenses because of this. all subjects without the sims were added and they all are processed similar to the other subjects
just fyi e.g. SA11 ASW00099dy, SA52 ASW0009ad3, SA68 ASW0009b7u and so on..