Soft Computing: (1997) 1 : (c) Springer-Verlag 1997

Editorial

What is soft computing?

Since the publication of my first paper on soft data analysis in 1981, 
the concept of soft computing has undergone many changes. In its
latest incarnation, soft computing may be defined as follows.
	Soft Computing (SC) is an association of computing 
methodologies centering on fuzzy logic (FL), neurocomputing (NC), 
genetic computing (GC), and probabilistic computing (PC). The 
methodologies comprising soft computing are for the most part 
complementary and synergistic rather than competitive.
	The guiding principle of soft computing is: exploit 
the tolerance for impression, uncertainty, partial truth, and 
approximation to achieve tractability, robustness, low solution cost 
and better rapport with reality. One of the principle aims of soft 
computing is to provide a foundation for the conception, design, and 
application of intelligent systems employing its member methodologies 
symbolically rather than in isolation.
	Within soft computing, the main concerns of fuzzy logic, 
neurocomputing, genetic computing and probabilistic computing center on:
	FL: approximate reasoning, information granulation, computing 
with words,
	NC: learning, adaptation, classification, system modelling 
and identification,
	GC: synthesis, tuning and optimization through systematized
random search and evolution,
	PC: management of uncertainty, belief networks, prediction,
chaotic systems.
	As an association of computing methodologies, soft computing 
is certain to grow in visibility and importance in the years ahead. 
What is the rationale behind this expectation? In my view, a key reason 
is related to the growing realization that the conceptual structure of 
conventional, hard computing is much too precise in relation to the 
pervasive imprecision of the real world.
	In this context, there are two distinct issues that have to be 
considered. First, there are many real world problems which do not 
lend themselves to solution by the techniques of hard computing because 
the need information is not available and/or the systems under 
consideration are not sufficiently well defined. Such problems are the 
norm in economic planning, living systems, large-scale societal systems 
and human decision-making. Another source of such problems is AI, 
especially in the realms of commonsense reasoning, computer vision and 
natural language understanding. Indeed, it may be argued that it is the 
commitment of mainstream AI to hard computing and its coolness toward 
soft computing that impeded AI's ability to achieve the ambitious goals 
that were set at its inception.
	The other and perhaps more important reason is that employment 
of soft computing methodologies serves to exploit the tolerance for 
imprecision, uncertainty, partial truth and approximation. In so doing, 
soft computing mimics the remarkable human ability to make rational 
decisions in an environment of uncertainty and imprecision. A case in 
point is the problem of parking a car. The tolerance for imprecision in 
this problem makes it possible for humans to park a car without any 
measurement and any knowledge of system dynamics. Without exploiting 
the tolarance for imprecision, the parking problem becomes intractable 
for humans and very hard for machines.
	Exploitation of the tolerance for imprecision, uncertainty, 
partial truth and approximation plays a pivotal role in data compression, 
information retrieval and communication. In this realm, fuzzy logic plays 
a particularly important role by providing a methodology for dealing with 
information granulation and computing with words in ways that mimic human 
reasoning and concept formation. In essence, the role model for fuzzy 
logic is the human mind.
	An aspect of soft computing that is of central importance is the 
symbolic relationship between its constituent methodologies. What this 
implies is that in the solution of many problems-especially in the conception
and design of intelligent systems-it is advantageous to employ a combination 
of two or more of the constituent methodologies of soft computing, leading to 
what is referred to as HYBRID INTELLIGENT SYSTEMS. Currently, the most visible 
systems of this type are neuro-fuzzy systems. However, we are also beginning
to see a growing number of fuzzy-genetic, neuro-genetic and neuro-fuzzy-genetic 
systems. Such systems are likely to become ubiquitous in the not distant future. 
What is certain is that-in the years ahead-the advent of hybrid intelligent 
systems will have a profound impact on the ways in which intelligent systems 
are conceived, designed, employed and interacted with.
	Viewed in this perspective, the publication of Soft Computing is an 
important event in the crystallization of soft computing as a prominent component 
of modern science and information technology. The publication of Soft Computing 
realizes the vision of the editors, the authors and the publisher-a vision which 
could not become a reality without the invaluable initiative and support of the 
SGS-THOMSON Corporation.


						Lotfi A. Zadeh
						Editor-in-chief