Article Preview
TopKecman provides a detailed insight of learning from experimental data and soft computing. Soft computing mimics the intelligent mathematical structures, or models, that underlie learning found in the nature (Kecman, 2001). According to Wang and Fu, the data dimensionality reduction (DDR) can reduce the dimensionality of hypothesis search space, reduce data collection and storage costs, enhance data mining performance and simplify the data mining result. They describe the use of the Genetic algorithm for feature selections as well (Wang & Fu, 2005). Zhang, Wang & Lin propose a unique subspace learning structure, Conjunctive Patches Subspace Learning (CPSL) process by employing the user historical feedback log data for a collaborative image retrieval (CIR) task. The CPSL can assimilate the discriminative information of labeled log images, the geometrical information of labeled log images and the weakly similar information (Zhang, Wang, & Lin, 2012). In addition, they have proposed a CIR for the reduction of labeling works of the user by resorting to the ancillary data. The support vector machine (SVM) vibrant learning can decide vague samples as the most instructive and thus, lessens the labeling works of conventional RF (Zhang, Wang, & Lin, 2014). Chen, Su, Gimson, Liu & Shine propose an object segmentation method to extract the image features, and showed that new feature types can be incorporated into the algorithm to further improve the image retrieval performance. They suggest that content-based image retrieval (CBIR) can be carried out to evaluate the object segmentation capability in dealing with the large-scale database images (Chen, Su, Gimson, Liu & Shine, 2012).