Rogério Rodrigues de Vargas

Ensembles of ARTMAP-based neural networks: an experimental study

ARTMAP-based models are neural networks which use a match-based learning procedure. The main advantage of ARTMAP-based models over error-based models, such as Multi-Layer Perceptron, is the learning time, which is considered as significantly fast. This feature is extremely important in complex systems that require the use of several models, such as ensembles or committees, since they produce robust and fast classifiers. Subsequently, some extensions of the ARTMAP model have been proposed, such as: ARTMAP-IC, RePART, among others. Aiming to add an extra contribution to ARTMAP context, this paper presents an analysis of ARTMAP-based models in ensemble systems. As a result of this analysis, two main goals are aimed, which are: to analyze the influence of the RePART model in ensemble systems and to detect any relation between diversity and accuracy in ensemble systems in order to use this relation in the design of these systems.

Interval ckMeans: An algorithm for clustering symbolic data

Clustering is the process of organizing a collection of patterns into groups based on their similarities. Fuzzy clustering techniques aim at finding groups to which every object in the database belongs to some membership degree. This paper presents a new algorithm for clustering symbolic data based on ckMeans algorithm. This new algorithm allows the data entry and the membership degree to be intervals. In order to validate the proposal, it is compared to two other algorithms using the same database.

A Comparison between K-Means, FCM and ckMeans Algorithms

Fuzzy C-Means, introduced by Jim Bezdek in 1981 is one of the earliest and most popular fuzzy clustering algorithms. However, in order to improve the hit rate or speed, over the years several modifications have been proposed. Among these we highlight the ckMeans algorithm proposed by us in 2010 which make a change in the way to calculate the center of the clusters of FCM. The idea is to use an auxiliary membership function of elements to those clusters that are essentially crisp and calculate the centroids following a similar process as done in K-Means algorithm but keeping the same procedures as in FCM in the rest of algorithm. In fact, this hybridization between FCM and K-Means motivated the name ckMeans for this variant of the FCM. In this article we apply K-Means, FCM and ckMeans algorithms in a validated database of mammograms with about a thousand elements and compare these three algorithms in terms of hit rate and number of each iterations and the computational processing time until the convergency of the system.

Identifying Pixels Classified Uncertainties ckMeansImage Algorithm

Floods may occur in rivers when the flow rate exceeds the capacity of the river channel, particularly at bends or meanders in the waterway. Floods often cause damage to homes and businesses becoming the most prevalent type of disaster in the world and the one with the highest number of events, causing the greatest economic losses, affecting a large number of people. This paper has the objective of mapping and identifying the flooding areas of a selected region in the municipality of Itaqui-RS using remote sensing. In order to do it, we used the Fuzzy ckMeansImage Algorithm to group and to classify the image into similarity clusters. The methodology consists in processing satellite images before and after the flooding occurs. Finally, we discuss the processed images and present the flooded area.

Using ckMeans algorithm in image segmentation process: Preliminary results on mammography analysis

Clustering algorithms aim at modelling fuzzy (i.e., ambiguous) unlabelled pat- terns efficiently. Our goal is to propose a ckMeans algorithm to the image segmentation process. To validate the proposed methodology we applied the algorithm in mammography images. We present the initial results considering just one image.