I am learning some machine learning (I have previous background in statistics) on my own, and I am interested in getting some hands-on practice by doing some ML analysis in R and Python.

Given how mad I am about astronomy, I would love to get my hands dirty on doing some ML using open data. I am aware that organisations like NASA have lots of open datasets. What I am less clear about is the kind of analysis that needs to be done.

So, I'd like to know if there is any website where I can get a repository of space data (images, signals, whatever) and which asks what analysis is required; kind of like Kaggle where one gets the dataset and has the required question which one can attempt.

  • $\begingroup$ You might have a look at some courses for statistics or astrostatistics first. That should channel your 'madness' into constructive and instructive channels. $\endgroup$ – AtmosphericPrisonEscape Dec 21 '18 at 0:40
  • 1
    $\begingroup$ My suggestion is that you should go to arXiv.org, and pick a paper that interests you. Then, you reproduce the analysis, so that you know what data, what methods, and what to aim for. $\endgroup$ – Kornpob Bhirombhakdi Dec 21 '18 at 2:22
  • $\begingroup$ @KornpobBhirombhakdi Unfortunately, that is often much easier said than done, many papers are not very clear when it comes to describing their analysis in detail. There are a few examples of papers out there that also include the full analysis stack, including Jupyter notebooks etc, but it is way too rare. $\endgroup$ – Thriveth Dec 24 '18 at 0:36
  • $\begingroup$ @Thriveth, I agree. However, you should have something to hang on and go from there. This strategy is still the best in my opinion, and I always suggest students to start with. While students can ask their professors in schools, a community like this one is important to support others like self learners or homeschool students when they need helps. $\endgroup$ – Kornpob Bhirombhakdi Dec 24 '18 at 13:53

A possibility might be a course at a university, youtube, or a self-study-course via a book.

What we've used in our astrostatistics class was "Wall & Jenkins: Practical Statistics for Astronomers" which presents theoretical sections with exercises as follow-ups. No answer sections exist, as you need to do some programing for some of them.

But you can maybe lend this or similar books at a local university library, without the need to spend money.

  • $\begingroup$ Thank you! I really appreciate it. I have lifetime membership of my University's library - will search there. If I don't find it there, then I don't mind buying the book if it is not too expensive. $\endgroup$ – Arpan Ganguli Dec 23 '18 at 23:15

The Python AstroML module might be interesting to you. It is accompanied by a textbook about machine learning applied to astronomical datasets, but if you don't have the cash to spend on the book, there are some exercises and datasets included in the module itself that should give you a lot to dig into.

  • $\begingroup$ Any idea what fraction of AstroML is Linux dependent. when I tried last time it seemed a lot of it needed Linux. $\endgroup$ – qqqqq Nov 21 '19 at 23:44

Here is a good source of public data https://pds-imaging.jpl.nasa.gov/ . My fear is that without some basic astronomy courses, there will be many little hassles ( e.g.1. the coordinates expressed in spherical coordinates etc, and you might expect xyz. FITS format reader )for any new person. Be mentally ready for numerous small obstacles.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.