What a great idea! I hope @George Varnavides gives you the info you need. I am curious, @Radek Pior, what type of planetarium you own? Perhaps you can give more details on equipment, size, maybe some images?... I sometimes visit Poland, it would be great to visit and see the Yoga Constellations on display!

I am thrilled that you shared this work George! Such an interesting synthesis of computing techniques!

Glad to hear you've enjoyed the project Hernik! I must admit I may not be the best person to answer your question - but I can provide some links which helped me, I had also never used/studied ML before Mathematica rolled them out in version 11.1.

1. For implementation of NN functionality in the WL, The Neural Networks Documentation Guide is VERY extensive. A nice summary of possible task-types can be seen under Applications on the NetTrain man page.
2. For importing state-of-the-art architectures, have a look at the nicely curated Wolfram NN Repository page.
3. For understanding some of the concepts better, I would suggest the easy-to-follow blog by Google's Christopher Olah (I particularly enjoyed his Recurrent NNs section).
4. Another nice summary blog (this one focuses on Convolutional NNs). In-fact using Spatial Transformer (implemented in WL) layers to identify constellation rotations is another direction one could go with this.
5. Finally if you ascribe to a Feynman-esque* method of learning, I would suggest coding a simple perceptron in WL to solve a linearly-separable problem and a back-propagation algorithm for a small network to solve a non-linearly separable problem.

Cheers, George

*Richard Feynman quote:

What I cannot create, I do not understand.

- Congratulations! This post is now a Staff Pick as distinguished by a badge on your profile! Thank you, keep it coming!