Want to understand generalization?

[TO BE WRITTEN]

• this one’s really hard.
• the question is basically: what functions can a neural network learn well with \(n\) samples?
• honestly, people don’t even try this empirically all that much. (open Q?)
• linear model / kernel / RF theory
• single- / multi-index model theory