The paper introduces an incremental learning method for the optimal construction of rule-based granular systems from numerical data streams. The method is developed within a multiobjective optimization framework considering the specificity of information, model compactness, and variability and granular coverage of the data.
COBISS.SI-ID: 12460116
In the paper, a new approach to evolving fuzzy model identification from streaming data is given. The structure of the model is given as a local model network in Takagi-Sugeno form, and the partitioning of the input-output space is based on metrics in which these local models are defined as prototypes of the clusters. The algorithm adds the local models in an incremental fashion and recursively adapts the local model parameters.
COBISS.SI-ID: 12504660
In this article, a new dynamic merging approach for incrementally evolving clustering is presented. The newly generated merged cluster is conducted by using the weighted averaging of cluster centers and the calculation of the joint covariance matrix from the covariance matrices of the clusters. It has been shown that the proposed new evolving algorithm eGAUSS+ together with the new merging concept is very easy to implement, can work on higher-dimensional data sets, can perform all necessary computation on-line, and can produce reliable clusters.
COBISS.SI-ID: 12586580
This article presents a novel method for fuzzy space partitioning and the identification of Takagi-Sugeno fuzzy models. The novelty is in its region-splitting mechanism and membership function definition, which is based on hyperplanes. The obtained results are very promising; however, as with most learning methods, the results depend on the data distribution and input variable selection.
COBISS.SI-ID: 12591444
In this paper, an efficient indoor localization algorithm based on the confidence-interval fuzzy model is presented. For each beacon in the test room, a new confidence-interval fuzzy path-loss model composed of several local linear models is constructed. By their consideration, the localization accuracy is significantly improved in comparison with other commonly used path-loss models.
COBISS.SI-ID: 12043092