Efficient Training Techniques for Classification with Vast Input Space

Donald C. Wunsch; Emad W. Saad; J. J. Choi; J. L. Vian

doi:10.1109/IJCNN.1999.831156

Efficient Training Techniques for Classification with Vast Input Space

Donald C. Wunsch
, Emad W. Saad
, J. J. Choi
, J. L. Vian

Research output: Contribution to journal › Article › peer-review

Abstract

Strategies to efficiently train a neural network for an aerospace problem with a large multidimensional input space are developed and demonstrated. The neural network provides classification for over 100,000,000 data points. A query-based strategy is used that initiates training using a small input set, and then augments the set in multiple stages to include important data around the network decision boundary. Neural network inversion and oracle query are used to generate the additional data, jitter is added to the query data to improve the results, and an extended Kalman filter algorithm is used for training. A causality index is discussed as a means to reduce the dimensionality of the problem based on the relative importance of the inputs

Original language	American English
Journal	Proceedings of the International Joint Conference on Neural Networks, 1999. IJCNN '99
DOIs	https://doi.org/10.1109/IJCNN.1999.831156
State	Published - Jan 1 1999

Keywords

Aerospace Problem
Causality Index
Classification
Computational Complexity
Dimensionality Reduction
Efficient Training Techniques
Extended Kalman Filter Algorithm
Filtering Theory
Jitter
Kalman Filters
Learning (Artificial Intelligence)
Multidimensional Input Space
Network Decision Boundary
Neural Nets
Neural Network
Neural Network Inversion
Oracle Query
Pattern Classification
Query-Based Strategy

Disciplines

Electrical and Computer Engineering

Access to Document

10.1109/IJCNN.1999.831156License: Unspecified

Efficient Training Techniques for Classification with Vast InputFinal published version, 544 KBLicense: Unspecified

Cite this

@article{54b7d0f126f94d32aa8a626c63fe7413,

title = "Efficient Training Techniques for Classification with Vast Input Space",

abstract = " Strategies to efficiently train a neural network for an aerospace problem with a large multidimensional input space are developed and demonstrated. The neural network provides classification for over 100,000,000 data points. A query-based strategy is used that initiates training using a small input set, and then augments the set in multiple stages to include important data around the network decision boundary. Neural network inversion and oracle query are used to generate the additional data, jitter is added to the query data to improve the results, and an extended Kalman filter algorithm is used for training. A causality index is discussed as a means to reduce the dimensionality of the problem based on the relative importance of the inputs",

keywords = "Aerospace Problem, Causality Index, Classification, Computational Complexity, Dimensionality Reduction, Efficient Training Techniques, Extended Kalman Filter Algorithm, Filtering Theory, Jitter, Kalman Filters, Learning (Artificial Intelligence), Multidimensional Input Space, Network Decision Boundary, Neural Nets, Neural Network, Neural Network Inversion, Oracle Query, Pattern Classification, Query-Based Strategy",

author = "Wunsch, \{Donald C.\} and Saad, \{Emad W.\} and Choi, \{J. J.\} and Vian, \{J. L.\}",

year = "1999",

month = jan,

day = "1",

doi = "10.1109/IJCNN.1999.831156",

language = "American English",

journal = "Proceedings of the International Joint Conference on Neural Networks, 1999. IJCNN '99",

issn = "2161-4407",

}

TY - JOUR

T1 - Efficient Training Techniques for Classification with Vast Input Space

AU - Wunsch, Donald C.

AU - Saad, Emad W.

AU - Choi, J. J.

AU - Vian, J. L.

PY - 1999/1/1

Y1 - 1999/1/1

N2 - Strategies to efficiently train a neural network for an aerospace problem with a large multidimensional input space are developed and demonstrated. The neural network provides classification for over 100,000,000 data points. A query-based strategy is used that initiates training using a small input set, and then augments the set in multiple stages to include important data around the network decision boundary. Neural network inversion and oracle query are used to generate the additional data, jitter is added to the query data to improve the results, and an extended Kalman filter algorithm is used for training. A causality index is discussed as a means to reduce the dimensionality of the problem based on the relative importance of the inputs

AB - Strategies to efficiently train a neural network for an aerospace problem with a large multidimensional input space are developed and demonstrated. The neural network provides classification for over 100,000,000 data points. A query-based strategy is used that initiates training using a small input set, and then augments the set in multiple stages to include important data around the network decision boundary. Neural network inversion and oracle query are used to generate the additional data, jitter is added to the query data to improve the results, and an extended Kalman filter algorithm is used for training. A causality index is discussed as a means to reduce the dimensionality of the problem based on the relative importance of the inputs

KW - Aerospace Problem

KW - Causality Index

KW - Classification

KW - Computational Complexity

KW - Dimensionality Reduction

KW - Efficient Training Techniques

KW - Extended Kalman Filter Algorithm

KW - Filtering Theory

KW - Jitter

KW - Kalman Filters

KW - Learning (Artificial Intelligence)

KW - Multidimensional Input Space

KW - Network Decision Boundary

KW - Neural Nets

KW - Neural Network

KW - Neural Network Inversion

KW - Oracle Query

KW - Pattern Classification

KW - Query-Based Strategy

UR - https://scholarsmine.mst.edu/ele_comeng_facwork/1000

U2 - 10.1109/IJCNN.1999.831156

DO - 10.1109/IJCNN.1999.831156

M3 - Article

SN - 2161-4407

JO - Proceedings of the International Joint Conference on Neural Networks, 1999. IJCNN '99

JF - Proceedings of the International Joint Conference on Neural Networks, 1999. IJCNN '99

ER -

Efficient Training Techniques for Classification with Vast Input Space

Abstract

Keywords

Disciplines

Access to Document

Other files and links

Fingerprint

Cite this