Overall, the article would benefit from including a more thorough description of the code as well as more detailed examples of application. I have noted a few suggestions that I list below.
The authors have designed the code for multi-wavelength astronomical time series, but, in the core of the paper, they do not really describe why multi-wavelength observations are relevant to study their objects of interest. In particular, why do we need GPR to model these light curves. For example, there is no mention of the modelling of stellar variability signals (e.g. activity), for example in the search for exoplanets (transit photometry or Doppler spectroscopy). This could be a highly relevant application of multi-wavelength GP modelling, since stellar activity signals are wavelength dependent.
The authors do not really demonstrate why pgmuvi is particularly adapted for multi-wavelength observations. What are the assumptions of model? A reminder of the mathematical framework would also be appreciated, even if this includes briefly describing GpyTorch. More importantly, I was not convinced by the comparison with other GPR codes (lines 65 to 80). This part lacks a more quantitative comparison with other codes (e.g. computation time and complexity of algorithm). The authors mention that similar results could be obtained with tinygp. Could the authors describe, in more details, (i) for which type of problems pgmuvi should be preferred and (ii) why the learning curve is easier in pgmuvi?
As said above, the applications of the code are limited to projects that have not been published (nor even submitted). Since the tutorials are limited to 1D- sine-wave fitting examples, the validation of the code on real data sets is not yet demonstrated. Could the authors please provide detailed examples of multi-wavelength data sets modelling with pgmuvi (if possible in the paper and in the tutorials – see the section about the Docs).
More minor comments:
Line 13, “Time-domain observations are increasingly important in astronomy”: What do the authors mean here? I believe that time-domain observations have occupied a central role in modern astronomy. The volume of data is indeed becoming increasingly important, but I believe that time-domain observations have been crucial for at least 50 years.
line 76: “Whic” → “Which”
I've created one issue to track all of these items. Feel free to tackle any of these if you want to, just tag your interest here! I'm assigning myself, but you can also add yourself if you're interested.
In the review (https://github.com/ICSM/pgmuvi/issues/45), @baptklein requested a few updates to the paper:
I've created one issue to track all of these items. Feel free to tackle any of these if you want to, just tag your interest here! I'm assigning myself, but you can also add yourself if you're interested.