Article Preview
TopIntroduction
Clustering is found to be one of the thrust research areas in computer vision and pattern recognition. There are various techniques and approaches to perform clustering in the literature (Pavel Berkhin, 2002; M. Steinbach, G. Karypis, V. Kumar, 2000). The reason for gaining high attention by the researchers is its wide variety of applications viz., segmentation, object recognition and content based information retrieval etc. In unsupervised approach, the relevant information search is found to be a difficult task (Cai, Deng and Zhang, Chiyuan and He, Xiaofei, 2010). For this, a Multi Cluster Feature Selection (MCFS) approach is used for solving the combinatorial optimization problem. It also solves the L1 regularized least squares and sparse Eigen problems. An incremental procedure is adopted for the K-means clustering method (AristidisLikas, NikosVlassis, JakobJ. Verbeek, 2003). It appends cluster by cluster dynamically. For this, it uses a globally deterministic search algorithm. With this, the computational load is also greatly reduced. The clustering techniques are found to be efficient to segment the vessels (Gehad Hassan, et al, 2015) in the input retina image. The K-means clustering algorithm and the mathematical morphology are used for smoothing and segmenting the image.
The inter neurons are clustered by using neuro morphological approach (Ivan Grbatinić, Nebojša Milošević, Bojana Krstonošić, 2017). It estimates the predictor and then performs the analysis of multivariate clusters and is then classified into multiple labels. To improve the efficiency of the Fuzzy Min-Max network, an ensemble of clustering tree (Manjeevan Seera, Kuldeep Randhawa, Chee Peng Lim, 2017) is used. It uses an efficient learning model to improve the performance of the clustering method. It uses four different types of mean measures in the network. A multi layer immune network (Xiao Yu, 2017) is used for clustering the fuzzy mid-wave infrared image. It uses the mechanism of the coordination network. It minimizes the inter class variance with immune neural network. The Euclidean distance based clustering with ART2 network is found to yield low performance when compared to other networks. To improve this, a steganographic projection with hybrid neural system (Andrzej Bielecki, Mateusz Wójcik, 2017) is used. It also improves the performance of both ART2 and RBF networks. It is observed that an unsupervised neural network (Marina Resta, Michele Sonnessa, Elena Tànfani, Angela Testi, 2017) is found to be efficient for analyzing the patterns of patient data flow information. In this, SOM is used for analyzing the data with clustering techniques and then the cluster information is represented with seasonal connotation. The failing banks (Michael Negnevitsky, 2017) are efficiently clustered with the SOM neural network. This method is found to be efficient to represent the likelihood of bank failures.