skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Adaptive Learning Theory.

Abstract

Abstract not provided.

Authors:
; ; ; ; ; ;
Publication Date:
Research Org.:
Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
Sponsoring Org.:
USDOE National Nuclear Security Administration (NNSA)
OSTI Identifier:
1367220
Report Number(s):
SAND2017-5192D
653358
DOE Contract Number:
AC04-94AL85000
Resource Type:
Conference
Resource Relation:
Conference: Proposed for presentation at the Poster display in Neural Exploration and Research Lab.
Country of Publication:
United States
Language:
English

Citation Formats

Vineyard, Craig Michael, Parekh, Ojas D., Phillips, Cynthia A., Aimone, James Bradley, James, Conrad D., Vineyard, Craig Michael, and Vineyard, Craig Michael. Adaptive Learning Theory.. United States: N. p., 2017. Web.
Vineyard, Craig Michael, Parekh, Ojas D., Phillips, Cynthia A., Aimone, James Bradley, James, Conrad D., Vineyard, Craig Michael, & Vineyard, Craig Michael. Adaptive Learning Theory.. United States.
Vineyard, Craig Michael, Parekh, Ojas D., Phillips, Cynthia A., Aimone, James Bradley, James, Conrad D., Vineyard, Craig Michael, and Vineyard, Craig Michael. 2017. "Adaptive Learning Theory.". United States. doi:. https://www.osti.gov/servlets/purl/1367220.
@article{osti_1367220,
title = {Adaptive Learning Theory.},
author = {Vineyard, Craig Michael and Parekh, Ojas D. and Phillips, Cynthia A. and Aimone, James Bradley and James, Conrad D. and Vineyard, Craig Michael and Vineyard, Craig Michael},
abstractNote = {Abstract not provided.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = 2017,
month = 5
}

Conference:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share:
  • Abstract not provided.
  • The authors describe an enhancement of the ARINC system testability and maintenance program (STAMP) to provide fault-isolation strategies that change and conform to evolving failure data. The enhancement is designed to produce an online fault-isolation strategy that learns through repeated applications. Three major areas are discussed. The first is an overview of the topological dependency-analysis methods implemented in STAMP, including design for testability. The second describes the development of fault-isolation strategies, including the basis for information-theory strategies. It also addresses the weighting of information theoretics to obtain certain fault-isolation objectives. The weighting algorithms can be used to optimize test costsmore » or test times, or to incorporate data on mean time between failures. The third section describes the techniques needed to provide information feedback for learning from previous applications of the interactive system. Examples are employed to illustrate the techniques. 9 references.« less
  • In this paper we propose and analyze a class of algorithms for neural networks training, which attempt to retain some of the most favourable features of both the online and the batch approach to backpropagation. More specifically, we consider a class of algorithms, where a suitable acceptance rule is introduced, which imposes periodically an upper bound on the learning rates, in a way that a sufficient decrease of the total error function is enforced. Given this bound, we perform a sequence of iterations based on the conjugate gradient method, by minimizing each of the error terms associated with the input-outputmore » patterns. Global convergence towards stationary points is established and a preliminary numerical comparison with online and batch backpropagation techniques is reported.« less
  • No abstract prepared.