19
$\begingroup$

What are some open problems in astronomy that an amateur would have a chance of solving? Suppose the amateur has a PhD in some other field, owns a basic telescope, a set of filters, diffraction gratings, cameras, and happens to know a lot about machine learning, signal processing, spectral estimation, statistics & design of experiments, and basic physics and chemistry.

Are there better tags available (such as "research")?

Improve this question
$\endgroup$
7
  • 1
    $\begingroup$ How does the discovery of new asteroids/comets/minor planets/etc affect the gravitational balance of the solar system, in particular NASA's projections of planetary locations and rotation rates. In 100 years, will there be only minor changes to the SPICE kernels, or extensive one? $\endgroup$
    – user21
    Commented Apr 5, 2015 at 14:00
  • $\begingroup$ Interesting question. Why not elaborate a little more on your comment @barrycarter and make it an answer? Since the question is about amateurs, it would be nice to know what kind of data would be needed and how it could be obtained. I'm also curious to hear about other open problems as well.. Does not astronomy have a plenty of them? :) $\endgroup$
    – mmh
    Commented Apr 6, 2015 at 16:48
  • 1
    $\begingroup$ I would strongly suggest going for low surface brightness objects: e.g. following these people: dunlap.utoronto.ca/instrumentation/dragonfly $\endgroup$
    – chris
    Commented Apr 6, 2015 at 17:10
  • 1
    $\begingroup$ An alternative would be to dig in surveys such as SDSS, DES etc.. using machine learning techniques to identify outliers. Many surveys are now publicly available and there is only so much professional astronomers can do. $\endgroup$
    – chris
    Commented Apr 6, 2015 at 17:11
  • $\begingroup$ I normally save my best ideas for my own grant proposals :) $\endgroup$
    – ProfRob
    Commented Apr 9, 2015 at 17:44

2 Answers 2

Reset to default
6
+50
$\begingroup$

If you have good knowledge of software development and pattern recognition, there are several problems that you could assist in solving. Much of observational astronomy requires long time series data, and removing the noise from this data. I have just left the field where some colleagues are trying to develop some software to use image subtraction techniques to isolate individual stars in the centre of clusters. The centre of the cluster is typically more dense and harder to get clear measurements of each individual star to analyse.

Pattern recognition would be particularly useful in pipeline analysis, where a generic pipeline is used on large amounts of data to 1: find the types of stars one is interested in; and 2: extracting some interesting information on these stars. Machine learning techniques could also be used to assist in developing the general pipelines for more specific interests.

I am happy to put you in touch with a few people that could provide you with some specific problems that you might be able to assist in.

Improve this answer
$\endgroup$
3
  • $\begingroup$ Is the offer also open to others than the OP? :) I might be interested too.. $\endgroup$
    – mmh
    Commented Apr 13, 2015 at 12:02
  • $\begingroup$ So if you would have for example this image upload.wikimedia.org/wikipedia/commons/thumb/0/02/…, the task would be to find coordinates of each star in center? :) An example would be nice. $\endgroup$
    – mmh
    Commented Apr 14, 2015 at 13:55
  • 1
    $\begingroup$ More along the lines of thousands of those images taken as a time series and then using common features in the images to 1:align them, and remove any features introduced by the detectors (heat drift, alignment error, etc); and 2: isolating the pixels associated with each star and determining the relative intensity for each. This gets difficult towards the centre where the intensity of the pixels are influenced by surrounding pixels. Here are some references which might help arxiv.org/abs/1009.4206, arxiv.org/abs/1309.6044 $\endgroup$
    – theotheraussie
    Commented Apr 21, 2015 at 4:14
5
$\begingroup$

The Kaggle galaxy zoo challenge is an example of a problem begging for ideas from outside the field. Sander Dieleman, with a background in deep learning and feature learning, bravely stepped forward, creating an image classifier utilising convolutional neural networks; his full solution is described fluently here.

These kinds of techniques could be applied to any image classification problem in Astronomy, or similar techniques could be used to classify other astrophysical objects from survey or signal data.

I would steer clear of doing your own image capture, as there are plenty of openly available data sets with greater depth, resolution and coverage than you could hope to carry out yourself within a reasonable time frame.

Improve this answer
$\endgroup$
1
  • $\begingroup$ I liked this answer too. :) $\endgroup$
    – mmh
    Commented Apr 16, 2015 at 19:27

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .