4 ms·
Is this a complicated way of saying "just use k-nearest neighbors"?
by isoprophlex 4y ago
Is this a complicated way of saying "just use k-nearest neighbors"?
- MrMan 4y agoI think its implying that ML often works because the manifold assumption is partially correct. so samples will lurk in the same neighborhood.
- crabbygrabby 4y agoIn a lot of cases knn doesn't model systems appropriately unless you understand the space very well. At least knn offers some sort of theoretical assumption of the bayes error rate though so it's a decent tool.