I have a lot of 3-dimensional data (x-value, y-value, and a given efficiency value z). Together, they form a heat map/efficiency map. These maps are almost similar; they just have small differences in the efficiency values.
Now, I want to train a model with this map data as output. After training, the model should be able to predict the entire map given just a few input vectors (x, y, z).
What would be the best model? Should I use a CNN, or would it be a generative task?
I tried using a GAN, but I was not able to condition it.