Aspects in Classification Learning - Review of Recent Developments in Learning Vector Quantization

Abstract

Abstract. Classification is one of the most frequent tasks in machine learning. However, the variety of classification tasks as well as classifier methods is huge. Thus the question is coming up: which classifier is suitable for a given problem or how can we utilize a certain classifier model for different tasks in classification learning. This paper focuses on learning vector quantization classifiers as one of the most intuitive prototype based classification models. Recent extensions and modifications of the basic learning vector quantization algorithm, which are proposed in the last years, are highlighted and also discussed in relation to particular classification task scenarios like imbalanced and/or incomplete data, prior data knowledge, classification guarantees or adaptive data metrics for optimal classification.

Precision-Recall-Optimization in Learning Vector Quantization Classifiers for Improved Medical Classification Systems

“…LVQ can be seen as a robust alternative for the k-nearest-neighbor classifier (k-NN, [27]- [29]) in the following sense [23], [30]: Assuming training samples v ∈ V ⊆ R n with class labels…”

“…Originally, LVQ was proposed to optimize classification accuracy and Bayes error based on Hebbian learning [22]. LVQ-models were proven to be powerful approaches for classification of vectorial and non-vectorial data [23]- [25].…”

Median variants of learning vector quantization for learning of dissimilarity data

“…Although original LVQ has been introduced on somewhat heuristic grounds [41], recent developments in this context provide a solid mathematical derivation of its generalization ability and learning dynamics [37]: LVQ classifiers can be substantiated by large margin generalization bounds [13,61,27]; the dynamics of LVQ type algorithms can be derived from explicit cost functions [63]. Interestingly, already the dynamics of classical LVQ provably leads to very good generalization ability in typical model situation as investigated in the framework of online learning [9].…”

Multivariate Stochastic Approximation to Tune Neural Network Hyperparameters for Criticial Infrastructure Communication Device Identification

“…LVQ algorithms are epistemologically self-organizing ANNs [37], and employ nearest neighbor approaches through the nearest prototype vector optimization process whereby the "nodes" or "prototype vectors" (PVs) of the ANN are iteratively moved to characterize the data [38]. In operation, LVQ algorithms train prototype vectors to a given class label by moving correctly classified PVs closer to a given class, incorrectly classified PVs are moved away from a given class.…”