Training Algorithm with Incomplete Data for Feed-ForwardNeural Networks

Authors:
Song-Yee Yoon;Soo-Young Lee
Affiliations:
Computation and Neural Systems Laboratory, Department of Electrical Engineering, Korea Advanced Institute of Science and Technology, 373–1 Kusong-dong, Yusong-gu, Taejon 305–701, Korea ...;Computation and Neural Systems Laboratory, Department of Electrical Engineering, Korea Advanced Institute of Science and Technology, 373–1 Kusong-dong, Yusong-gu, Taejon 305–701, Korea ...
Venue:
Neural Processing Letters
Year:
1999

Citing 5
Cited 9

Optimal brain damage

Advances in neural information processing systems 2
Generalization by weight-elimination with application to forecasting

NIPS-3 Proceedings of the 1990 conference on Advances in neural information processing systems 3
Hierarchical mixtures of experts and the EM algorithm

Neural Computation
Information geometry of the EM and em algorithms for neural networks

Neural Networks
Merging back-propagation and Hebbian learning rules for robust classifications

Neural Networks

Using Unlabelled Data to Train a Multilayer Perceptron

Neural Processing Letters
A Hybrid Neural Network System for Pattern Classification Tasks with Missing Features

IEEE Transactions on Pattern Analysis and Machine Intelligence
Learning from imperfect data

Applied Soft Computing
A study on the use of imputation methods for experimentation with Radial Basis Function Network classifiers handling missing attribute values: The good synergy between RBFNs and EventCovering method

Neural Networks
Learn++.MF: A random subspace approach for the missing feature problem

Pattern Recognition
Missing value imputation on missing completely at random data using multilayer perceptrons

Neural Networks
Soft computing based imputation and hybrid data and text mining: The case of predicting the severity of phishing alerts

Expert Systems with Applications: An International Journal
Change Detection of Remote Sensing Images with Semi-supervised Multilayer Perceptron

Fundamenta Informaticae
Classifying patterns with missing values using Multi-Task Learning perceptrons

Expert Systems with Applications: An International Journal

Quantified Score

Hi-index	0.00

Visualization

Abstract

A new algorithm is developed totrain feed-forward neural networks for non-linearinput-to-output mappings with small incomplete data inarbitrary distributions. The developedTraining-EStimation-Training (TEST) algorithm consistsof 3 steps, i.e., (1) training with the completeportion of the training data set, (2) estimation ofthe missing attributes with the trained neuralnetworks, and (3) re-training the neural networks withthe whole data set. Error back propagation is stillapplicable to estimate the missing attributes. Unlikeother training methods with missing data, it does notassume data distribution models which may not beappropriate for small training data. The developedTEST algorithm is first tested for the Iris benchmarkdata. By randomly removing some attributes from thecomplete data set and estimating the values latter,accuracy of the TEST algorithm is demonstrated. Thenit is applied to the Diabetes benchmark data, of whichabout 50% contains missing attributes. Compared withother existing algorithms, the proposed TEST algorithmresults in much better recognition accuracy for testdata.