Improved Adaptive Splitting and Selection: The Hybrid Training Method of a Classifier Based on a Feature Space Partitioning

Abstract

Currently, methods of combined classification are the focus of intense research. A properly designed group of combined classifiers exploiting knowledge gathered in a pool of elementary classifiers can successfully outperform a single classifier. There are two essential issues to consider when creating combined classifiers: how to establish the most comprehensive pool and how to design a fusion model that allows for taking full advantage of the collected knowledge. In this work, we address the issues and propose an AdaSS+, training algorithm dedicated for the compound classifier system that effectively exploits local specialization of the elementary classifiers. An effective training procedure consists of two phases. The first phase detects the classifier competencies and adjusts the respective fusion parameters. The second phase boosts classification accuracy by elevating the degree of local specialization. The quality of the proposed algorithms are evaluated on the basis of a wide range of computer experiments that show that AdaSS+ can outperform the original method and several reference classifiers.

Discrimination of Schizophrenia Auditory Hallucinators by Machine Learning of Resting-State Functional MRI

“…The application of machine learning to multiple imaging modalities rather than to rsfMRI data alone 69 would also help to confirm our results. In addition, classification could be further enhanced by innovative unsupervised 56; 60 or supervised 4; 39 approaches. Finally, while the brain localizations derived from features with strong classification performance provide information about brain areas that likely have biological relevance in AH pathophysiology, the discriminant voxels do not contain information about the directionality of connectivity (i.e., hyper- or hypoconnectivity) with LHG and RHG.…”

Ensemble classification for imbalanced data based on feature space partitioning and hybrid metaheuristics

“…-To address the last two tasks (feature space centroids and individual weights optimization for each partition and ensemble, respectively) into one integrated process. This approach has proven to obtain very good results in other works as [29,30,31]. The main novelty of our proposal w.r.t.…”

“…Concretely, we focus on the approach called AdaSS (Adaptive Splitting and Selection) that simultaneously divides the feature space into partitions and establishes a different classifier for each partition. This approach was proposed by [29] and recently used in other works as [30,31] with very promising results. This approach for building ensembles entails the resolution of a complex optimization problem whose objective is the minimization of the error of the whole system.…”

Fuzzy rule based classification systems for big data with MapReduce: granularity analysis

“…An additional possibility consists of the combination of the RBs into an ensemble classifier . This way, we can also take advantage of a wider space partitioning carried out in the Map stage (Jackowski et al 2014;Wozniak and Krawczyk 2012).…”