Contemporary knowledge discovery, data mining, and other systems for intelligent data processing comprise a rapidly developing domain with well-established standards. In spite of a number of fundamental results and practical applications, the domain is far from being in a finished-up state; many fundamental questions still are far from being resolved. Amid them, questions of vital importance concerning the computational complexity of the applied algorithms, combining algorithmic and data-driven approaches, result convergence, effective use of statistical and logical methods and so on are still to be resolved. The purpose of this special issue is to present various points of view and different research that will show different dimensions of one and the same domain: intelligent data processing. The idea is that the reader himself/herself to discover repeating elements in the presented heterogeneous methods and achievements and to determine priority trends and perspectives in the domain: How have systems become more and more intelligent and the cooperation between human-machine or machine-machine become more and more creative. Five papers from this domain are offered and these are as follows.
Visual novelty detection and tracking systems are considered in the paper "An Approach to Automatic Real-Time Novelty Detection, Object Identification, and Tracking in Video Streams Based on Recursive Density Estimation and Evolving Takagi-Sugeno Fuzzy Systems" by P. Angelov, P. Sadeghi-Tehran, and R. Ramezani (UK). Their computational efficient approach has evolving property; it is more effective than most of the other contemporary systems, and it has less demand to memory storage costs. Also it does not require human operator in the loop. The main idea of the approach is to approximate the probability distribution function of the color intensity by a Cauchy-type kernel and to use a recursive expression to update this estimation online by the information from the pixel color intensity that the next image frame brings. The proposed approach is faster by 10 times or more than the well-known kernel density estimation methods. If combined with real-time prediction methods, a fast and fully autonomous system can be realized with potential applications in surveillance and robotic systems.