1
$\begingroup$

I'm trying to work with Calcium-K line profiles from the Sun.

Image for reference. Please ignore the labels on the image and note that my profiles are not in image format (more info below).

Calcium-K line profile

I have over 1500 profiles like these for various latitudes and dates. The profiles are saved as 2D numpy arrays, where each array has two rows: the first row contains the wavelength values, and the second row contains the intensity values. So, plotting against each other, we get the profiles as shown above.

I would like to use an ML model to train latitude and time with line profiles. So, given future date and latitude, the model generates a possible line profile.

I'm stuck with how I can map latitude and time with the line profiles. If I flatten each numpy array and make a list of 1D numpy arrays, ANN, LSTM, and CNN models totally fail and show garbage. Random Forest, XGBoost, and KNN don't fare well either.

What algorithm can I use to solve this? And more importantly, how do I retain line profile info (that is lost if flattened) while training?

Any pointers would help.

Improve this question
$\endgroup$
5
  • 5
    $\begingroup$ Data Science may help. This really isn't a physics question. $\endgroup$ Commented Jul 14, 2023 at 13:50
  • 3
    $\begingroup$ For reference, Jon is talking about our sister site, Data Science . $\endgroup$ Commented Jul 14, 2023 at 14:04
  • 1
    $\begingroup$ @JonCuster, yes, I figured as much and posted there as well. I just wanted to see if anyone in the physics community here has attempted this. Ultimately, whatever I get from it, I will have to understand the physics of it, so. But thanks nonetheless. $\endgroup$ Commented Jul 14, 2023 at 15:25
  • $\begingroup$ ...I will have to understand the physics of it... What physics? You're asking about methods of storing information for use in a program/function which is purely CS. $\endgroup$ Commented Jul 14, 2023 at 18:20
  • 1
    $\begingroup$ @KyleKanos, I'm referring to what I get after I'm done training the model and forecasting with it. My thinking was that since many physicists have started using ML more these days, I could check if anyone has worked on anything similar. $\endgroup$ Commented Jul 14, 2023 at 18:27

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.