A General Probabilistic Formulation for Supervised Neural Classifiers

Authors:
Hongmei Ni;Tülay Adali;Bo Wang;Xiao Liu
Affiliations:
Department of Computer Science and Electrical Engineering, University of Maryland Baltimore County, Baltimore, MD 21250, USA;Department of Computer Science and Electrical Engineering, University of Maryland Baltimore County, Baltimore, MD 21250, USA;Department of Computer Science and Electrical Engineering, University of Maryland Baltimore County, Baltimore, MD 21250, USA;GlobeSpan Semiconductor Inc., 100 Schulz Drive, Red Bank, NJ 07701, USA
Venue:
Journal of VLSI Signal Processing Systems
Year:
2000

Citing 7
Cited 1

An introduction to signal detection and estimation (2nd ed.)

An introduction to signal detection and estimation (2nd ed.)
Information geometry of the EM and em algorithms for neural networks

Neural Networks
The theoretical and experimental status of the n-tuple classifier

Neural Networks
Neural Networks: A Comprehensive Foundation

Neural Networks: A Comprehensive Foundation
Partial likelihood for estimation of multi-class posterior probabilities

ICASSP '99 Proceedings of the Acoustics, Speech, and Signal Processing, 1999. on 1999 IEEE International Conference - Volume 02
Conditional distribution learning with neural networks and itsapplication to channel equalization

IEEE Transactions on Signal Processing
Quantification and segmentation of brain tissues from MR images: a probabilistic neural network approach

IEEE Transactions on Image Processing

Partial likelihood for online order selection

Signal Processing - Special issue: Information theoretic signal processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

We use partial likelihood (PL) theory to introduce a general probabilistic framework for the design and analysis of neural classifiers. The formulation allows for the training samples used in the design to have correlations in time, and for use of a wide range of neural network probability models including recurrent structures. We use PL theory to establish a fundamental information-theoretic connection, show the equivalence of likelihood maximization and relative entropy minimization, without making the common assumptions of independent training samples and true distribution information. We use this result to construct the information geometry of partial likelihood and derive the information geometric e- and m-projection (em) algorithm for class conditional density modeling by finite normal mixtures. We demonstrate the successful application of the algorithm by a channel equalization example and give simulation results to show the efficiency of the scheme.