ASYMPTOTIC AND FINITE-SAMPLE PROPERTIES OF ESTIMATORS BASED ON STOCHASTIC GRADIENTS
ASYMPTOTIC AND FINITE-SAMPLE PROPERTIES OF ESTIMATORS BASED ON STOCHASTIC GRADIENTS
About this item
Full title
Author / Creator
Publisher
Hayward: Institute of Mathematical Statistics
Journal title
Language
English
Formats
Publication information
Publisher
Hayward: Institute of Mathematical Statistics
Subjects
More information
Scope and Contents
Contents
Stochastic gradient descent procedures have gained popularity for parameter estimation from large data sets. However, their statistical properties are not well understood, in theory. And in practice, avoiding numerical instability requires careful tuning of key parameters. Here, we introduce implicit stochastic gradient descent procedures, which in...
Alternative Titles
Full title
ASYMPTOTIC AND FINITE-SAMPLE PROPERTIES OF ESTIMATORS BASED ON STOCHASTIC GRADIENTS
Authors, Artists and Contributors
Author / Creator
Identifiers
Primary Identifiers
Record Identifier
TN_cdi_proquest_journals_1927469091
Permalink
https://devfeature-collection.sl.nsw.gov.au/record/TN_cdi_proquest_journals_1927469091
Other Identifiers
ISSN
0090-5364
E-ISSN
2168-8966
DOI
10.1214/16-AOS1506
