On the Optimal Memorization Power of ReLU Neural Networks

Gal Vardi · Gilad Yehudai · Ohad Shamir


Keywords: [ Memorization ] [ deep learning theory ] [ theory ]

[ Abstract ]
[ Visit Poster at Spot D2 in Virtual World ] [ Slides [ OpenReview
Wed 27 Apr 2:30 a.m. PDT — 4:30 a.m. PDT
Spotlight presentation:

Abstract: We study the memorization power of feedforward ReLU neural networks. We show that such networks can memorize any $N$ points that satisfy a mild separability assumption using $\tilde{O}\left(\sqrt{N}\right)$ parameters. Known VC-dimension upper bounds imply that memorizing $N$ samples requires $\Omega(\sqrt{N})$ parameters, and hence our construction is optimal up to logarithmic factors. We also give a generalized construction for networks with depth bounded by $1 \leq L \leq \sqrt{N}$, for memorizing $N$ samples using $\tilde{O}(N/L)$ parameters. This bound is also optimal up to logarithmic factors. Our construction uses weights with large bit complexity. We prove that having such a large bit complexity is both necessary and sufficient for memorization with a sub-linear number of parameters.

Chat is not available.