Login

Publications  •  Project Statistics

Glossary  •  Schools  •  Disciplines
People Search: 
   
Title/Abstract Search: 

Dissertation Information for Bambang Parmanto

NAME:
- Bambang Parmanto

DEGREE:
- Ph.D.

DISCIPLINE:
- Library and Information Science

SCHOOL:
- University of Pittsburgh (USA) (1995)

ADVISORS:
- Paul Munro

COMMITTEE MEMBERS:
- Robert Korfhage
- Marek J. Druzdzel
- Alice Smith
- Howard Doyle

MPACT Status: Fully Complete

Title: Agitating dissent: Methods for improving performance of a neural network committee by error decorrelation

Abstract: The idea of a neural network committee is to combine several neural network predictors, instead of using a single network, to perform collective decision-making such as majority voting, simple averaging, or weighted averaging. The potential of a committee in improving the prediction performance has been well documented. Central to the performance improvement of the committee compared to the individual networks is the error correlation between networks in the committee. Methods of achieving error independence between the networks are the subject of investigation in this dissertation. The proposed methods have two major bulwarks. The first is to use resampling techniques to compose different learning set replicas for each network in the committee. The second is learning algorithms based on ancillary tasks to decorrelate the networks. The effectiveness of the proposed methods are demonstrated in data sets with different levels of noise and in real-world medical diagnosis problems.

MPACT Scores for Bambang Parmanto

A = 0
C = 3
A+C = 3
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2010-09-28 00:22:18

Advisors and Advisees Graph

generating graph, please reload

Students under Bambang Parmanto

ADVISEES:
- None

COMMITTEESHIPS:
- Ming Mao - University of Pittsburgh (2008)
- Qiang Ye - University of Pittsburgh (2010)
- Yefei Peng - University of Pittsburgh (2010)