CountSketches, Feature Hashing and the Median of Three

Publikation: Bidrag til bog/antologi/rapportKonferencebidrag i proceedingsForskningfagfællebedømt

Dokumenter

  • Fulltext

    Forlagets udgivne version, 450 KB, PDF-dokument

In this paper, we revisit the classic CountSketch method, which is a sparse, random projection that transforms a (high-dimensional) Euclidean vector v to a vector of dimension (2t - 1)s, where t, s > 0 are integer parameters. It is known that a CountSketch allows estimating coordinates of v with variance bounded by parallel to v parallel to/s. For t > 1, the estimator takes the median of 2t - 1 independent estimates, and the probability that the estimate is off by more than 2 parallel to v parallel to/root s is exponentially small in t. This suggests choosing t to be logarithmic in a desired inverse failure probability. However, implementations of CountSketch often use a small, constant t. Previous work only predicts a constant factor improvement in this setting. Our main contribution is a new analysis of CountSketch, showing an improvement in variance to O(min{parallel to v parallel to, parallel to v parallel to/s}) when t > 1. That is, the variance decreases proportionally to s, asymptotically for large enough s.

OriginalsprogEngelsk
TitelProceedings of the 38 th International Conference on Machine Learning
RedaktørerM Meila, T Zhang
ForlagPMLR
Publikationsdato2021
Sider6011-6020
StatusUdgivet - 2021
Begivenhed38th International Conference on Machine Learning (ICML) - Virtual
Varighed: 18 jul. 202124 jul. 2021

Konference

Konference38th International Conference on Machine Learning (ICML)
ByVirtual
Periode18/07/202124/07/2021
NavnProceedings of Machine Learning Research
Vol/bind139
ISSN2640-3498

Links

ID: 301135523