An Organized Committee of Artificial Neural Networks in the Classification of Human Chromosomes

Abstract

Neural networks are organized in committees to improve the correctness of the decisions created by artificial neural networks (ANN's). In the classification of human chromosomes, it is accustomed to use multilayer perceptrons with multiple (22-24) outputs. Because of the huge number of synaptic weights to be tuned, these classifiers cannot go beyond a level of 92% overall correctness. In this study we represent a special organized committee of 462 simple perceptrons to improve the rate of correct classification of 22 types of human chromosomes. Each of these simple perceptrons is trained to distinguish between two types of chromosomes. When a new data is entered, the votes of these 462 simple perceptrons and additional 22 dummy perceptrons create a decision matrix of the size 22×22. By a special assembling of these votes we get a higher rate of correct classification of 22 types of human chromosomes.

Metric learning for sequences in relational LVQ

“…From the database we use the "CopChromTwo" subset for binary classification, containing classes 4 and 5 with 200 sequences each (N = 400). In the literature, these two classes have been reported to yield a lower recognition rate than the others, see [10]. The authors in [10] used an organized ensemble of multilayer perceptrons to classify all 22 chromosomes in the Copenhagen database, and list the classification accuracies for individual classes.…”

“…In the literature, these two classes have been reported to yield a lower recognition rate than the others, see [10]. The authors in [10] used an organized ensemble of multilayer perceptrons to classify all 22 chromosomes in the Copenhagen database, and list the classification accuracies for individual classes. For the chromosomes 4 and 5, they report 91% and 89% accuracy on the test set, respectively, whereas the overall average is 95.86%.…”

Human chromosome classification using Competitive Neural Network Teams (CNNT) and Nearest Neighbor

“…The recent work of Gagula-Palalic and Can [18] has shown that classification using CNNTs, tested on Copenhagen chromosome data set gives error rate less than 5%. In this method, each expert of committee of neural networks is trained to distinguish two classes only, unlike the previous ANN based methods, where neural networks are trained to classify more than two classes, if not all, at the same time [1], [12], and [13].…”

“…In this method, each expert of committee of neural networks is trained to distinguish two classes only, unlike the previous ANN based methods, where neural networks are trained to classify more than two classes, if not all, at the same time [1], [12], and [13]. In usual committee machines, the voting process is based on majority vote, boosting, or weighted votes, whereas in [18], the final outcome of the testing data is a decision matrix, containing outcomes of each machine and the classes are determined using a special voting scheme, as described in the section III-A.…”