Generalization Error Bounds
One theoretical motivation behind margin classifiers is that their generalization error may be bound by parameters of the algorithm and a margin term. An example of such a bound is for the AdaBoost algorithm. Let be a set of examples sampled independently at random from a distribution . Assume the VC-dimension of the underlying base classifier is and . Then with probability we have the bound
for all .
Read more about this topic: Margin Classifier
Famous quotes containing the words error and/or bounds:
“They have their belief, these poor Tibet people, that Providence sends down always an Incarnation of Himself into every generation. At bottom some belief in a kind of pope! At bottom still better, a belief that there is a Greatest Man; that he is discoverable; that, once discovered, we ought to treat him with an obedience which knows no bounds. This is the truth of Grand Lamaism; the “discoverability” is the only error here.”
—Thomas Carlyle (1795–1881)
“How far men go for the material of their houses! The inhabitants of the most civilized cities, in all ages, send into far, primitive forests, beyond the bounds of their civilization, where the moose and bear and savage dwell, for their pine boards for ordinary use. And, on the other hand, the savage soon receives from cities iron arrow-points, hatchets, and guns, to point his savageness with.”
—Henry David Thoreau (1817–1862)