Supervised subspace projections for constructing ensembles of classifiers

A novel Boosted-neural network ensemble for modeling multi-target regression problems

“…One approach to deal with complex, real-world problems is to combine AI prediction models and form an ensemble of predictors that exploit the different local behaviors of the base models to improve the overall prediction system's performance (Masoudnia et al, 2012). The main objective of ensemble learning methods is to simplify a difficult prediction task by dividing it into some relatively easy prediction subtasks and formulating a consensus prediction result for the original data (García-Pedrajas et al, 2012). From another perspective, ensemble learning is an approach to enhance the prediction accuracy for complex problems, such as those involving a limited number of instances, highdimensional feature sets, and highly complex trends and behaviors (Kotsiantis, 2011).…”

“…One of the causes of boosting failure is putting too much emphasis on correctly classifying all instances. Outliers, or noisy instances become too relevant in the training set, undermining the ensemble's performance (García-Pedrajas et al, 2012).…”

“…The notion of using subspace projection methods to construct NNE models has been applied in different works (Garc´ıa-Pedrajas et al, 2007;Garc´ıa-Pedrajas and Ortiz-Boyer, 2008). More recently, García-Pedrajas et al (2012) proposed a new boosting-based method for designing classifier ensembles according to unsupervised and supervised projection of random subspaces using different projection methods. The projections were constructed using only misclassified instances to focus the next classifier on the most difficult dataset instances.…”

Adaptive generalized ensemble construction with feature selection and its application in recommendation

“…Wei et al have proposed an ensemble approach by combining the predictions made by Positive Naïve Bayes with the classifier of positive example-based learning for the unlabeled examples [5]. García-Pedrajas et al have used the supervised projections of random subspaces to construct ensemble, which combines the philosophy of boosting to generate supervised projection based on the misclassified samples, and then trains using all available samples in the space given by the supervised projections [6]. Bock et al have adopted a statistical technique for nonparametric, the generalized additive model (GAM), as the ensemble component and proposed three GAM ensemble classifiers for binary classification based on Bagging, random subspace and a combination of both [7].…”

Incremental maximum margin criterion based on eigenvalue decomposition updating algorithm

“…Face recognition [1][2][3][4][5][6] has been an active research area in computer vision and pattern recognition communities in the last decades. Since two-dimensional face images are usu-B Gui-Fu Lu luguifu_tougao@163.com 1 School of Computer and Information, Anhui Polytechnic University, Wuhu 241000, Anhui, China ally transformed into one-dimensional vectors via column by column or row by row concatenation in face recognition, the original input-image space has a very high dimension and a dimensionality reduction technique is usually employed to solve the high-dimensionality problem.…”