The diversity/accuracy dilemma: An empirical analysis in the context of heterogeneous ensembles

Classifier ensembles for image identification using multi-objective Pareto features

“…Although a high degree of diversity is desired in gen-40 eral [15], it has also been found that diversity sometimes negatively impacts the overall ensemble accuracy, in particular when all base classifiers are weak [16,17]. In other words, to construct accurate base classifiers is as important as to create highly diverse base classifiers.…”

Bagging of Complementary Neural Networks with Double Dynamic Weight Averaging

“…Diversity in an ensemble system can be managed when the individual classified machines is created under different situations, which are different parameter settings of the classifiers, different classifier training datasets, and different classifier types [3]. In this paper, diversity can be reached by creating the individual classified machine under different classifier training datasets using bagging learning strategies with N components in an ensemble [2].…”

“…Beside the diversity, accuracy of individual classifiers is also an important consideration. To derive better performance of ensemble, the trade-off between diversity and accuracy have to be considered [3]. It was found that a diverse ensemble of less accurate classification can yield better performance than an ensemble of more accurate classification with less diversity [17].…”

“…Therefore, several researchers [5], [6], [7], [12], [13], [14], [15] have utilized the ensemble system to improve the accuracy on their works. One aspect that can improve the prediction performance of the ensemble technique is the system diversity [3]. Diversity is a measure that defines the disagreement degree in the output of the individual classified machines in the ensemble [11].…”