NATIONAL BUREAU OF ECONOMIC RESEARCH
NATIONAL BUREAU OF ECONOMIC RESEARCH

Generalized Stochastic Gradient Learning

George W. Evans, Seppo Honkapohja, Noah Williams

NBER Technical Working Paper No. 317*
Issued in October 2005
NBER Program(s):   TWP

We study the properties of generalized stochastic gradient (GSG) learning in forward-looking models. We examine how the conditions for stability of standard stochastic gradient (SG) learning both differ from and are related to E-stability, which governs stability under least squares learning. SG algorithms are sensitive to units of measurement and we show that there is a transformation of variables for which E-stability governs SG stability. GSG algorithms with constant gain have a deeper justification in terms of parameter drift, robustness and risk sensitivity.

You may purchase this paper on-line in .pdf format from SSRN.com ($5) for electronic delivery.

Information about Free Papers

You should expect a free download if you are a subscriber, a corporate associate of the NBER, a journalist, a site with your domain name in ".GOV", or a resident of nearly any developing country or transition economy.

If you usually get free papers at work/university but do not at home, you can either connect to your work VPN or proxy (if any) or elect to have a link to the paper emailed to your work email address below. The email address must be connected to a subscribing college, university, or other subscribing institution. Gmail and other free email addresses will not have access.

E-mail:

Machine-readable bibliographic record - MARC, RIS, BibTeX

 
Publications
Activities
Meetings
Data
People
About

National Bureau of Economic Research, 1050 Massachusetts Ave., Cambridge, MA 02138; 617-868-3900; email: info@nber.org