ABSTRACT
Soft computing (SC), in fact, greatly differs from the conventional (we can call it hard computing, or better firm computing! for a software [SW] or an algorithm/program resides in the HW) because it is tolerant of (i) imprecision, (ii) uncertainty, (iii) partial truth and (iv) approximation, and the role model for SC is as such the human mind, since the human mind (the humind) does not always take hard decisions [1], it takes soft or often vague decisions. The directive principle of SC is: exploit the tolerance for imprecision (of definitions/expressions of the views), uncertainty in partial decisions/data/ information, (available) partial truth and approximation (of computing new decisions/estimations) to achieve tractability/solvability, robustness and low cost of solutions [1]. The principal paradigms of SC are fuzzy logic (FL), neural computing (artificial neural networks [ANNs], NC), evolutionary computation (EC → GAs), machine learning (ML) and probabilistic reasoning (PR). Our biological neural networks (BNNs) and our many decision-making abilities work as if on the principles of imprecision and uncertainty. The belief networks (NWs) (D-S theory of decision making), chaos theory and parts of learning theory are also the supportive disciplines of some of these components that do or carry out the SC. Hence, the SC is a sharing partnership in which each one of the components contributes a distinct method with its own features and the characteristic way of addressing a problem, in totality, in its domain by using the available information/data that are partially imprecise, uncertain and incomplete. Our BNNs and our (humans) decisionmaking approaches also always try to make sense out of such data and in such environments, and in such cases we hardly use formal mathematics, be it learning, adaptation and/or data fusion (DF). Thus, SC paradigms in fact then, model our behaviours and activities and also provide computational tools so that we can formally and legitimately use these techniques for solving difficult problems and even develop and build artificial intelligent (AI) systems that work like humans. Such AI-based systems HW/SW/algorithms/procedures) can also work in hazardous situations and areas, and can also perform tasks of data/image processing, decision making and DF.
