This document introduces soft computing and provides an agenda for the lecture. Soft computing is defined as a fusion of fuzzy logic, neural networks, evolutionary computing, and probabilistic computing to deal with uncertainty and imprecision. Hybrid systems combine different soft computing techniques for improved performance. The lecture will cover an introduction to soft computing, fuzzy computing, neural networks, evolutionary computing, and hybrid systems. References are also provided.
2. Agenda
• Introduction of softcomputing
• Course outline
• Recap of neural networks
The student already familiar with neural network may
leave after the introduction of softcomputing
3. Introduction (1/3)
What is Softcomputing ?
• The idea of softcomputing was initiated in 1981 when Lofti A.Zadeh
published his first paper on soft data analysis “what is
softcomputing”, softcomputing. Springer-Verlag Germany/ USA, 1997.
• Zedeh, define softcomputing into one multidisciplinary system as the
fusion of the fields of Fuzzy Logic, Neuro-computing, Evolutionary
computing and Probabilistic Computing.
• An essential aspect of soft computing is that its constituent
methodologies are, for the most part, complementary and symbiotic
rather than competitive and exclusive.
• Softcomputing breakdown
SC = EC + NC + FL + PC
Softcomputing Evolutionary Neural Fuzzy Logic Probabilistic
4. Introduction (2/3)
What is meant by fusion or hybridization ?
• Hybridization create a situation where different
entities cooperate advantageously for final
outcome
• For example, EC can be employed in the design of
fuzzy-logic-based systems to improve or optimize
their performance. In the reverse direction, the
machinery of fuzzy logic can be employed to
improve the performance of genetic algorithms.
5. Introduction (3/3)
• Currently, the most visible systems of this
hybrid type are Neuro-Fuzzy (NF) systems,
Fuzzy-Genetic (FG) systems, Neural-Genetic
(NG) systems, Fuzzy-Neural-Genetic (FNG)
systems, Fuzzy-Probablistic (FP) systems.
Other combinations are also possible.
• So we are not concerned with EC, FL and NN
in isolation (as in AI, ML) but hybridization is
the prime concern here.
6. Primary Role of Individual Constituents
in the Hybridization (1/2)
The core of SC consist of several paradigms mainly: neural
computing, evolutionary computing, probabilistic computing
and fuzzy systems.
• Neural computing: the importance of neurocomputing
derives in large measure from the fact that NC provides
effective algorithms for the purpose of system
identification, classification, learning and adaptation.
• Evolutionary computing: The primary contribution of
evolutionary computing is a machinery for systematic random
search. Such search is usually directed at finding an optimum
solution to a problem. Genetic algorithms and modes of
genetic computing, e.g., genetic programming, may be viewed
as special cases of evolutionary computing.
7. Primary Role of Individual Constituents
in the Hybridization (2/2)
• Probabilistic computing: the primary contribution
of probabilistic computing is the machinery of
probability theory and the subsidiary techniques for
decision-making under uncertainty.
• Fuzzy logic: the primary contribution of fuzzy logic
is the machinery of knowledge representation via
fuzzy if-then rules and to perform logic inference like
FOL with the ability to handle uncertainty and
imprecision.
8. Hard Computing (Classical Soft Computing
Artificial Intelligence) Soft (Computational Intelligence)
Hard Vs Computing
Prime desiderata is precision and certainty. It is Exploit tolerance for imprecision and
traditional AI which is based on two principles: uncertainty. The aim is to model the
firstly, represent knowledge in symbolic form remarkable abilities of human mind which
(i.e. Letters, words, phrases, signs). Secondly, characteristically exploit the tolerance for
search the solution with the aid of symbolic imprecision and uncertainty to e.g. understand
logic (e.g. FOL). Despite success of AI for the distorted speech, sloppy handwritten,
developing numerous applications (e.g. Expert expressions in natural language and drive a
systems, natural language understanding, vehicle in dense traffic, etc
theorem proving). It is enable to deal with
advance requirement such as speech
recognition, hardwritten recognition,
computer vision, machine translation, learning
with experience
Require programs to be written Can evolve its own programs
Deterministic Stochastic
Require exact input Can deal with ambiguous and noisy data
Produce precise answer Produce approximate answers
Table: Listed in the table are some differences between hard and soft computing. The
list is not exhaustive.
9. Structure of Soft Computing
Computing Methodologies
Computing Methodologies Computing Methodologies
Fuzzy Systems Neural Computing
Soft Computing: Hybrid
Systems or Fused System
Probabilistic Computing Evolutionary Computing
10. Definition
Lofti A. Zedah, 1992: “softcomputing is an emerging
approach to computing which parallel the
remarkable ability of human mind to reason and
learn in the environment of uncertainly and
imprecision”
11. Course Outline (1/2)
• Introduction
Definition, goals and importance; recap: fuzzy computing, neural
computing, genetic algorithm
• Fuzzy computing
Fuzzy computing: Classical set theory, crisp and non-crisp
set, capturing certainty, definition of fuzzy set; graphic
interpretations
• Neural Computing
Biological model, artificial neuron, architectures, learning
methods, Taxonomy of NN systems, single and multilayer
perceptrons, applications
• Evolutionary Computing
Genetic algorithms, taxonomy of optimization and evolution
techniques: guided random search techniques, calculus-based
techniques, genetic algorithms, evolutionary algorithms
12. Course Outline (2/2)
• Associative Memory
Description of AM, Examples of Auto and Hetro AM
• Adaptive Resonance Theory
Recap: supervised and unsupervised learning, back
propagation; competitive learning, stability and plasticity
dilemma, ART networks, Iterative clustering, Unsupervised
ART clustering
• Hybrid systems
Integration of neural network, fuzzy logic and genetic
algorithms, GA based back propagation network, fuzzy back
propagation network, fuzzy associative memories
13. References
• Zadeh L. A. Soft Computing and Fuzzy Logic. IEEE Software 11 (6): 48-
58, 1998.
• Lofti A.Zadeh. what is softcomputing”, softcomputing. Springer-Verlag
Germany/ USA, 1997.
• Rajasekaran S., G. A Vijayalaksmi Pai. Neural Network, Fuzzy Logic, and
Genetic Algorithms, Prentice Hall, 2005.
• K. Naresh, Sinha, M. Gupta. Soft Computing and Intelligent Systems –
Theory and Applications, Academic Press, 2000.
• Fahreddine Karray. Soft Computing and Intelligent System Design –
Theory, Tools and Applications, Addison Weslay, 2004.
• Tettamanzi, Andrea, Tomassine. Soft Computing: Integrating
Evolutionary, Neural and Fuzzy Systems, Springer, 2001.
• J. S. R Jang, C. T. Sun. Neuro-Fuzzy and SoftComputing: A Computational
Approach to Learning and Machine Intelligance, Prentice Hall, 1996.