I think no - SVMs are explicitly optimized to generalize the best from small data (https://en.wikipedia.org/wiki/Hinge_loss https://en.wikipedia.org/wiki/Hinge_

I think no - SVMs are explicitly optimized to generalize the best from small data (https://en.wikipedia.org/wiki/Hinge_loss https://en.wikipedia.org/wiki/Hinge_loss - 'The hinge loss is used for "maximum-margin" classification'), whereas NNs have more hacky regularization methods. I am not sure if the same is true for tree-based methods, but of course those are lovely due to how interpretable they are when you have few features.