Accurate and Fast Off and Online Fuzzy ARTMAP-Based Image Classification With Application to Genetic Abnormality Diagnosis

Abstract

Abstract-We propose and investigate the fuzzy ARTMAP neural network in off and online classification of fluorescence in situ hybridization image signals enabling clinical diagnosis of numerical genetic abnormalities. We evaluate the classification task (detecting a several abnormalities separately or simultaneously), classifier paradigm (monolithic or hierarchical), ordering strategy for the training patterns (averaging or voting), training mode (for one epoch, with validation or until completion) and model sensitivity to parameters. We find the fuzzy ARTMAP accurate in accomplishing both tasks requiring only very few training epochs. Also, selecting a training ordering by voting is more precise than if averaging over orderings. If trained for only one epoch, the fuzzy ARTMAP provides fast, yet stable and accurate learning as well as insensitivity to model complexity. Early stop of training using a validation set reduces the fuzzy ARTMAP complexity as for other machine learning models but cannot improve accuracy beyond that achieved when training is completed. Compared to other machine learning models, the fuzzy ARTMAP does not loose but gain accuracy when overtrained, although increasing its number of categories. Learned incrementally, the fuzzy ARTMAP reaches its ultimate accuracy very fast obtaining most of its data representation capability and accuracy by using only a few examples. Finally, the fuzzy ARTMAP accuracy for this domain is comparable with those of the multilayer perceptron and support vector machine and superior to those of the naive Bayesian and linear classifiers.Index Terms-Fluorescence in situ hybridization (FISH), fuzzy ARTMAP neural network (NN), genetic abnormality diagnosis, image classification, off-and online learning.

Predictive Abilities of Bayesian Regularization and Levenberg–Marquardt Algorithms in Artificial Neural Networks: A Comparative Empirical Study on Social Data

“…Preventing the model from defining a bias and choosing a statistical method that is robust and can solve complex relationships are also crucial. An Artificial Neural Network (ANN) is a popular statistical method which can explore the relationships between variables with high accuracy [1][2][3][4]. Essentially, the structure of an ANN is computer-based and consists of several simple processing elements operating in parallel [3,5,6].…”

Polytope ARTMAP: Pattern Classification Without Vigilance Based on General Geometry Categories

“…From elementary geometry, we know that every point is a linear combination of vectors in the set , so which verify (13) Thus, the set of vectors is linearly dependent, and they compose a singular matrix (14) where is the th component of . Equation (14) is the definition of in (3) (Section III-A1).…”

“…These networks feature online, incremental learning, with self-organized (ART1, ART2, ART3, fuzzy ART) and supervised learning (fuzzy ARTMAP (FAM) [2], Gaussian ARTMAP (GAM) [3], distributed ARTMAP (DAM) [4], ellipsoid ARTMAP (EAM) [5], FasArt [6], and FAM with relevance (FAMR) [7], among others). These models have been widely used in many application fields, including robotics [8], data mining [9], information fusion [10], data clustering [11], multichannel pattern recognition [12], image classification [13], etc. Default ARTMAP [14] compiles the basic structure of an ARTMAP network.…”

Advanced Developments and Applications of the Fuzzy ARTMAP Neural Network in Pattern Classification

“…FAM and its variants have exemplified themselves as accurate and fast learners in performing various classification tasks, such as automatic target recognition based on radar range profiles [8], speaker-independent vowel recognition [9], online handwritten recognition [10], QRS-wave recognition [11], medical diagnosis of breast cancer and heart disease [12], three-dimensional object understanding and prediction from a series of two-dimensional views [13], classification of noisy signals [14], discrimination of alcoholics from nonalcoholics [15] and recently genetic abnormality diagnosis [16], as well as many other classification tasks such as [17,18,19,20,21,22,23].…”

“…It is the self-organization nature of the algorithm that while enabling continuous learning of novel patterns also leads to the overfitting of noisy (overlapped) data that are mistakenly considered novel. This sensitivity can lead to uncontrolled growth in the number of categories, also referred to as category proliferation, leading to high computational and memory complexities and possible degradation in classification accuracy [7,9,16,24,25,26,27,28,29]. Improving FAM classification accuracy and reducing model category proliferation are the most studied topics in the FAM literature.…”