Meta-learning has 3 broad approaches: model, metric and optimization-based approach. Each of them has its own sub-approach, like matching network, meta-agonistic and Siamese-based network, and so on.
How do I decide which approach to select for a task? For my case, I have a noisy image, and they need to be compared with 10 different new images every time. Do I have to start with the trial and error method, or there is some methodology behind this approach selection?