Adaptation of fuzzy cognitive maps by migration algorithms

Abstract

Purpose -Conventional rule-based systems are insufficient for description of complex dynamic systems requiring nontrivial decision procedures. Fuzzy cognitive maps seem to be convenient to overcome these limitations. However, they lack ability of self-learning and therefore some adaptation approaches are needed. The purpose of this paper is both to show the use of fuzzy cognitive maps for such systems and to present migration algorithms as convenient adaptation means. Design/methodology/approach -Some problems of a complex dynamic system description by knowledge-based means are discussed. Fuzzy cognitive maps are presented as a possible way to solve these problems followed by description of migration algorithms as their adaptation means. Their use is clarified on an example of the so-called parking problem based on path planning using a graph search algorithm and a traffic simulation system. Findings -After series of simulations the reality of the proposed system and selected methods with their modifications was proved. It has shown the robustness of the presented solution under circumstances of uncertainty, too.Research limitations/implications -The paper points to stability investigation of the proposed approach introducing uncertainties into the traffic simulation system to take into account, e.g. unexpected events. Further, a possibility of developing a linguistic information retrieval system is mentioned. Practical implications -The proposed approach can find various implementations not only in planning tasks but also in robotic navigation and multi-agent applications in general. In addition, it suggests possibilities of knowledge-based systems, directly using human-like approaches, to areas of decision making under uncertainties and contradictories. Originality/value -An new modification of migration algorithms for adaptation of parameters for fuzzy cognitive maps is introduced and compared to other known self-learning methods. Further, the concept of a traffic simulation system for path planning is presented.

Local weather prediction system for a heating plant using cognitive approaches

“…They can be used for transforming the prediction results in a more comprehensive form for human perception [10,56], maybe later their personalization for a concrete user. Now we are experimenting with some adaptation approaches of knowledge bases for fuzzy systems [57,58].…”

Optimal behaviour prediction using a primitive-based data-driven model-free iterative learning control approach

“…The learning is expected to converge also under trial repeated disturbances and considering different o.f.s as well [37][38][39][40][41][42][43][44][45][46]. …”

Petri Nets to B-Language Transformation in Software Development

“…csp2B and CSP||B) will certainly be considered here. On the PN side the demonstrated approach can also benefit from cooperation with other formalisms, for example with the linear logic [20] or artificial intelligence concepts such as cognitive maps, including their fuzzy variants [24].…”