View Single Post
Old 04-21-2013, 07:56 AM
nkatz nkatz is offline
Junior Member
Join Date: Apr 2013
Posts: 4
Default Why 2^N Makes Learning Unfeasible

I was going to ask what was wrong with 2^N growth function since 2^Ne^{-N}\rightarrow 0 but I think I figured it out:
The right hand side of the modified Hoeffding is really 2^{2N}/e^{\epsilon^2N/8} which goes to 0 only if 2^2/e^{\epsilon^2/8}<1, which would require \epsilon>3.3 which is meaningless since Ein and Eout are between 0 and 1.
Is this argument correct?
Reply With Quote