My problem may not be very general. Imagine you observe the sun through thick clouds in multiple wavelengths. Given the spectrum of light starting off the sun's surface, you can calculate how rays will scatter, be absorbed and re-emit passing through the cloud given the 3d distribution, mass, and chemical composition of water vapors.
I want to do similar thing with my hydrodynamical galaxy simulation data. Classically, we've done this by following rays and it is time consuming. I believe if I need it scientifically accurate, I will have to solve the equations forward.
But, just for the visualization purpose, would there be a way to get rough synthetic images given 3d position and properties of light sources, scattering and absorbing material?
I am not an ML expert. So, I want to know if this exercise will worth anything before I start.
Generative Adversarial Networks (GANs) are a key example of unsupervised machine learning. The method involves a multi-level neural network generator which creates random guesses of the target image and at each step, the discriminator network seeks alternate neural network paths giving improved candidate images until the output becomes an exact replica. Quoting from the paper ‘A Beginner’s Guide To Generative Adversarial Networks (GANs)’ which is available on Google “the generator is trying to fool the discriminator while the discriminator is trying to not get fooled by the generator. As the models train through alternating optimization, both methods are improved until a point where the counterfeits are indistinguishable from the genuine ones.”
- Badges
- Science topic