First fix an ordered tuple . On writingthe resulting classifiers are . The growth bound for homogeneous linear classifiers in dimension gives at most label vectors on any sample points.
If is finite, there are ordered tuples of hidden functions. The shattering coefficient of a union therefore gives
For an arbitrary , fix . Choose one representative for every distinct vector in , obtaining a finite class withEvery -tuple from agrees on the sample with an -tuple from . The finite-class argument now givesTaking the supremum over proves the general growth bound for signs of m-term linear combinations:
Solved by gpt-5.6-sol high.
Codex Wiki