Talk:VSOP Variability Classification Subproject

I have changed the order of the discussion. Latest edits of this page are at the top, so to start writing go to the top of the page.

Light curves from NSVS for 5 objects in our set of stars were collected by Matilde and sent to Luis Manuel for a variability classification with the photometric classifier. The stars are: AP Ari, AL Leo, BP Psc, DM Boo and IM Vir. The results will be posted and discussed here.--Pedro 08:06, 4 January 2007 (CLST)

By cross-checking the list of more than 1000 stars (compare with the 110+ variability types) used in the training of the photometric classifier for COROT and the list of 200 stars observed up to date for VSOP, I have found an overlap of 3 stars. The stars are:


 * BT Psc
 * LP Vir
 * V342 Peg

Now, why these stars are in a list to train a variability-type classifier and in a list of stars of unknown variability type is a mistery which we should check and make sure that it has only been a relaxation of constraints on our behalf. Curiously the three of them are gamma Dor-type stars (SIMBAD) and may be this has something to do with them being in our list.


 * Pedro, just a quick comment: GCVS lists two of them as ELL and one as IB. From the CCF it is clear that at least two of them are binaries. In general I find SIMBAD very little reliable in terms of variability type (GCVS is also not perfect of course). -- Tdall 22:52, 1 December 2006 (CLST)


 * Thomas, these three stars have been included in the training set for the COROT classifier as gamma Dor (waiting for confirmation from Luis Manuel) and are well observed objects. At this point I trust the gamma Dor classification more than that of the GCVS. By the way, have we considered the effect of LPVs in the CCF? From the bibliography, spectroscopic data for at least two of these stars suggest single status. Another point, why are we observing stars with hi-res spectroscopic data already taken?.--Pedro 08:06, 4 January 2007 (CLST)

Next step: I will contact Luis Manuel and I will ask him about the probabilities that the COROT's classifier gives for the variability types of these stars and then compare with our spectra. I want to see if it gives, for instance, 100% probability of being gamma Dor variables or if there are also some other possibilities. The spectra for these three stars should look very similar (I can check it).

Regarding the stars with individual sections in PaperOne, I have asked Matilde to collect any existing light curves. They will be sent to Luis Manuel for a variability type probability determination. More news soon. --Pedro 20:32, 1 December 2006 (CLST)
 * The stars have been changed to those mentioned above. The stars in the sections of the paper did not have measured light curves.--Pedro 08:06, 4 January 2007 (CLST)

Another comment. I have just come from a talk given by the P.I. of the 2nd-generation, GTC-instrument SIDE which will have, among others, the capability of obtaining more than 300 spectra at a resolution of between 15000 and 20000. One of the scientific cases was that this instrument could do the follow-up of GAIA to obtain RV for stars for which GAIA will give no information. I imagine other instrumens providing MOS at medium to high resolution coming up (not counting the existing ones). With the tool we intend to develop with this subproject, there should be no problem in providing spectral/variability types and even physical parameters for the GAIA targets.--Pedro 08:45, 28 November 2006 (CLST)


 * Certainly very exciting prospects! What we need to do now is to identify the tasks that need to be done, and then organize the work.
 * "...only need line fluxes/EWs..."; which lines? same for all spectral types? how to measure that? what about rotational broadening? ...?
 * - The idea is to measure as many lines as posible in the spectrum. Then the final selection will be optimize between the groups working in the project (VSOP, Granada, Madrid and Leuven). I forsee also other parameters of the lines being needed, like higher order moments or bisectors. --Pedro 17:46, 29 November 2006 (CLST)
 * OK, this step actually requires then, that somebody (or some automatic tool!) actually measured the lines before feeding it into this classifier. And bisectors and moments as well: Bisectors can be done automatically, don't know about moments, but why not. Bottom line is, we need some other tool to actually measure the lines, before feeding the classifier. -- Tdall
 * "...several types of well known variables..."; which classes of variables? only secure variables? how many types? how broad a span in physics? ...?
 * - I got a list from Luis Manuel with all the objects used for the photometric training set and they are a lot. I am not sure how many of the variability types they represent. I guess that at least a single spectrum for all of them will be needed.--Pedro 17:47, 29 November 2006 (CLST)
 * Sounds reasonable. You can cross-check that list with GCVS. I guess there won't be many of them in VSOP, since the learning stars are presumably well-studied ones. Nevertheless, you should get the major types represented in the learning sample. -- Tdall
 * "...have a working package in a few weeks..."; doing what? how will it work? what are the exact requirements for it? what will/should it deliver? ...?
 * - Remember that they already have people working on this. I will have to talk more thorougly with Luis Manuel and the people in Leuven to see how they are tackling the problem.--Pedro 17:46, 29 November 2006 (CLST)
 * These are the kinds of questions we should consider carefully, then we should explore some design ideas (what we want it to do), and then we should make some precise requirements to hand over to Luis Manuel and the other (sub-)teams. I think it is more important that this project be done well, rather than some fractional quick-and-dirty solution gets hacked for Paper One.  So, stay with the long time-frame and do it well IMO. There will be other Papers...
 * I completely agree. I was just giving out another possibility in case things with paper one got stuck. -- Pedro
 * I understand. I'm just trying to play "the devils advocate" here. I'm trying to force you to define very carefully what this subproject is all about, what the goals are for VSOP, and what kind of work it involves. It may be very difficult for you to balance the requirements of the outside teams and the amount of commitment you'll be able to secure from within VSOP. Don't get me wrong; I think this project could be very important for us, since it's something that would be directly useful, and used by a lot of people for a lot of future science. -- Tdall


 * Also, Pedro, I'd like a little more status description on the projects of the other groups, i.e. the "CoRoT variability classifier" and "GAIAs WPs on (un)supervised classification of variability", and the "variability characterization of classes". Could you make a few subsections here explaining a bit about these projects, their aims and their status? -- Tdall 15:29, 28 November 2006 (CLST)
 * I too am very interested in knowing more about those projects. My intention is to attend the meetings and workshops on these topics. There is very little in the literature about them.--Pedro 20:32, 1 December 2006 (CLST)

I have talked to Luis Manuel once again and I have told him about our hurries for developing a spectroscopic classification tool. He says that to have a working (but not the best) tool he would only need line fluxes/EWs, considered by us as interesting to determine the variability/spectral type, from several types of well known variables. With this he would have a working package in a few weeks. The working line would be:

1. Classify a number of variable stars by determining spectral types and variability type. The variability type could be determined with the photometric light curve classifier and the light curves of these objects.

2. Produce a list of lines fluxes or EWs from which to classify spectral type and/or variability from these well classified variables.

This is a quick solution that can be implemented in a few weeks but it is not the optimum one. Perhaps it is already too much work for PaperOne. But after PaperOne's deadline, this is something that will have to be done anyway. I remember Cédric suggesting a compilation of lines, ratios of lines or EWs so that any of us could do the spectroscopic analysis for any star. Well, we are at this point again.

Regarding the variability typing of our objects for paper one, we can send them light curves and we can get probabilities of variability types in a few days. As suggested by Thomas, we (Matilde and I) will collect light curves for a few objects and send them to Luis Manuel to get these probabilities. Then, we will compare with the spectroscopic data.

Another thing to do for this project is to compare the list of the stars in their photometric training dataset and our objects. Ideally, spectra of the stars already in their training set would be collected if not already available. --Pedro 07:02, 28 November 2006 (CLST)