COMPUTATIONAL INTELLIGENCE tion (16).

Consider, for example, a collection of articial life

forms with behaviors resembling ants. Fitness will be mea-
sured relative to the total number of pieces of food found
and eaten (partially eaten food is counted). Reproduction
INTRODUCTION
consists in selecting the ttest individual x and weakest in-
dividual y in a population, and replacing y with a copy of x.
There are a number of interpretations of the notion of Com-
After reproduction, a population will then have two copies
putational Intelligence (CI) (19). Computationally intelli-
of the ttest individual. A crossover operation consists in
gent systems have been characterized by Bezdek (1, 2) rel-
exchanging genetic coding (bit values of one or more genes)
ative to adaptivity, fault-tolerance, speed, and error rates.
in two different chromosomes. The steps in a crossover op-
In its original conception, a number of technologies were
eration are (1) randomly select a location (also called inter-
identied to constitute the backbone of Computational In-
stitial location) between two bits in a chromosome string to
telligence, namely, neural networks (75, 76), genetic algo-
form two fragments, (2) select two parents (chromosomes
rithms (75, 76), fuzzy sets and fuzzy systems (75, 76), evo-
to be crossed), and (3) interchange the chromosome frag-
lutionary programming (75, 76) and articial life (10, 11).
ments. Because of the complexity of traits represented by
More recently, rough set theory and its extensions to ap-
a gene, substrings of bits in a chromosome are used to rep-
proximate reasoning and real-time decision systems have
resent a trait (17). The evolution of a population resulting
been considered in the context of computationally intelli-
from the application of genetic operations results in chang-
gent systems (3,69,12,13,46,75,76), which naturally led to
ing tness of individual population members. A principal
the generalization along the line of Granular Computing.
goal of GAs is to derive a population with optimal tness.
Overall, CI can be regarded as a eld of intelligent system
The pioneering works of Holland (15) and L. J. Fogel and
design and analysis which dwells upon a well-dened and
others (18) gave birth to the new paradigm of population-
clearly manifested synergy of genetic, granular and neural
driven computing (evolutionary computation) resulting in
computing. A detailed introduction to the different facets
structural and parametric optimization. Evolutionary pro-
of such a synergy along with a discussion of various real-
gramming was introduced by L. J. Fogel in the 1960s (19).
izations of such synergistic links between CI technologies
The evolution of competing algorithms denes evolution-
is given in (3,4,44,46,65,66,75,76).
ary programming. Each algorithm operates on a sequence
of symbols to produce an output symbol that is likely to
GENETIC ALGORITHMS maximize an algorithms performance relative to a well-
dened payoff function. Evolutionary programming is the
Genetic algorithms were proposed by Holland as a search precursor of genetic programming (15). In genetic program-
mechanism in articially adaptive populations (14). A ge- ming, large populations of computer programs are geneti-
netic algorithm (GA) is a problem-solving method that sim- cally bred.
ulates Darwinian evolutionary processes and naturally oc-
curring genetic operations on chromosomes (15). In nature,
a chromosome is a threadlike linear strand of DNA and as- FUZZY SETS AND SYSTEMS
sociated proteins in the nucleus of animal and plant cells. A
chromosome carries genes and serves as a vehicle in trans- A fuzzy systems (models) are immediate constructs that re-
mitting hereditary information. A gene is a hereditary unit sults from a description of real-world systems (say, social,
which occupies a specic location on a chromosome and economic, ecological, engineering, or biological) in terms
which determines a particular trait in an organism. Genes of information granules- fuzzy sets and relationships be-
can undergo mutation (alteration or structural change). A tween them (20). The concept of fuzzy set introduced by
consequence of the mutation of genes is the creation of a Zadeh in 1965 (21, 22) becomes of paramount relevance
new trait in an organism. In genetic algorithms, the traits when formalizing a notion of partial membership of ele-
of articial life forms are stored in bit strings which mimic ment. Fuzzy sets are distinguished from the fundamental
chromosome strings found in nature. The traits of individ- notion of a set (also called a crisp set) by the fact that their
uals in a population are represented by a set of evolving boundaries are formed by elements with whose degree of
chromosomes. A GA transforms a set of chromosomes to belongingness are allowed to assume numeric values in the
obtain the next generation of an evolving population. Such interval [0, 1]. Let us recall that the characteristic func-
transformations are the result of applying operations such tion for a set X returns a Boolean value {0, 1} indicating
as reproduction based on survival of the ttest and ge- whether an element x is in X or is excluded from it. A fuzzy
netic operations such as sexual recombination (also called set is non-crisp inasmuch as the characteristic function for
crossover) and mutation. a fuzzy set returns a value in [0, 1]. Let U, X, A, x be a
Each articial chromosome has an associated tness, universe of objects, subset of U, fuzzy set in U, and an in-
which is measured with a tness function. The simplest dividual object x in X, respectively. For a set X, A : X [0,
form of tness function is known as raw tness, which is 1] is a function which determines the degree of member-
some form of performance score (e.g., number of pieces of ship of an object x in X. A fuzzy set A is then dened to be
food found, amount of energy consumed, number of other a set of ordered pairs where A = {(x, A (x)) | x X}. The
life forms found). Each chromosome is assigned a proba- counterparts of intersection and union (crisp sets) are the
bility of reproduction which is proportional to its tness. t-norm and s-norm operators in fuzzy set theory. For the
In a Darwinian system, natural selection controls evolu- intersection of fuzzy sets, the min operator was suggested

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright 2007 John Wiley & Sons, Inc.
2 Computational Intelligence

by Zadeh (29), and belongs to a class of intersection opera- tion. Rough set theory also offers a model for approxima-
tors (min, product, bold intersection) known as triangular tion of vague concepts (69, 83).
or t-norms. A t-norm is a mapping t : [0, 1]2 [0, 1]. The In particular, rough set methods provide a means of ap-
s-norm (t-conorm) is a mapping s : [0, 1]2 [0, 1] (also tri- proximating a set by other sets (33, 34). For computational
angular co-norm) is commonly used for the union of fuzzy reasons, a syntactic representation of knowledge is pro-
sets. The properties of triangular norms are presented in vided by rough sets in the form of data tables. In general,
(82). an information system IS is represented by a pair (U, F),
Fuzzy sets exploit imprecision in conventional systems where U is a non-empty set of objects and F is a non-empty,
in an attempt to make system complexity manageable. It countable set of probe functions that are a source of mea-
has been observed that fuzzy set theory offers a new model surements associated with object features. For example, a
of vagueness (13). Many examples of fuzzy systems are feature of an image may be color with probe functions that
given in Pedrycz (23), and in Kruse, Gebhardt, and Kla- measure tristimulus values received from three primary
wonn (24). color sensors, brightness (luminous ux), hue (dominant
wavelength in a mixture of light waves), and saturation
(amount of white light mixed with a hue). Each f F maps
NEURAL COMPUTING
an object to some value. In effect, we have f : U Vf for
every f F.
Neural networks offer a powerful and distributed comput-
The notions of equivalence and equivalence class are
ing architecture equipped with signicant learning abili-
fundamental in rough sets theory. A binary relation R X
ties (predominantly as far as parametric learning is con-
X is an equivalence relation if it is reexive, symmetric
cerned). They help represent highly nonlinear and multi-
and transitive. A relation R is reexive if every object x X
variable relationships between system variables. Starting
has relation R to itself. That is, we can assert x R x. The sym-
from pioneering research of McCulloch and Pitts (25), and
metric property holds for relation R if xRy implies yRx for
others (26, 27), neural networks have undergone a signi-
every x, y X. The relation R is transitive for every x, y, z
cant metamorphosis and have become an important reser-
X, then xRy and yRz imply xRz. The equivalence class of an
voir of various learning methods (28) as well as an exten-
object x X consists of all objects y X so that xRy. For each
sion of conventional techniques in statistical pattern recog-
B A, there is associated an equivalence relation IndA (B)
nition (29). Articial Neural Networks (ANNs) were intro-
= {(x, x) | B. (x) = (x)} (indiscernibility relation). If (x,
duced to model features of the human nervous system (25).
x) IndA (B), we say that objects x and x are indiscernible
An articial neural network is collection of highly inter-
from each other relative to attributes from B. This is a fun-
connected processing elements called neurons. In ANNs, a
damental concept in rough sets. The notation [x]B is a com-
neuron is a threshold device, which aggregates (sums) its
monly used shorthand that denotes the equivalence class
weighted inputs, and applies an activation function to each
dened by x relative to a feature set B. In effect, [x]B = {y
aggregation to produce a response. The summing part of a
U | x IndA (B) y}. Further, partition U/IndA (B) denotes the
neuron in an ANN is called an Adaptive Linear Combiner
family of all equivalence classes of relation IndA (B) on U.
(ALC) in (30, 31). A McCulloch-Pitts neuron ni is a binary
Equivalence classes of the indiscernibility relation (called
threshold unit with
an ALC that computes a weighted sum
n B-granules generated by the set of features B (13)) repre-
net where net = j=0
w j x j . A weight wi associated with
sent granules of an elementary portion of knowledge we
xi represents the strength of connection of the input to a
are able to perceive relative to available data. Such a view
neuron. Input x0 represents a bias, which can thought of
of knowledge has led to the study of concept approxima-
as an input with weight 1. The response of a neuron can
tion (40) and pattern extraction (41). For X U, the set X
be computed in a number of ways. For example, the re-
can be approximated only from information contained in B
sponse of neuron ni can be computed using sgn(net), where
by constructing a B-lower and B-upper approximation de-
sgn(net) = 1 for net > 0, sgn(net) = 0 for net = 0, and sgn(net)
noted by B X = {x U | [x]B X} and B*X = {x U | [x]B X
= 1, if net < 0. A neuron comes with adaptive capabili-
= }, respectively. In other words, a lower approximation
ties that could be fully exploited assuming that there is an
B X of a set X is a collection of objects that can be classied
effective procedure is introduced to modify the strengths
with full certainty as members of X using the knowledge
of connections so that a correct response is obtained for a
represented by features in B. By contrast, an upper approx-
given input. A good discussion of learning algorithms for
imation B X of a set X is a collection of objects representing
various forms of neural networks can be found in Freeman
both certain and possible uncertain knowledge. In the case
and Skapura (32) and Bishop (29). Various forms of neural
where B X is a proper subset of B X, then the objects in X
networks have been successfully used in system modeling,
cannot be classied with certainty, and the set X is rough.
pattern recognition, robotics, and process control applica-
It has recently been observed by Pawlak (13) that this is
tions (46,50,51,54,75,76).
exactly the idea of vagueness proposed by Frege (41). That
is, the vagueness of a set stems from its borderline region.
ROUGH SETS The size of the difference between lower and upper ap-
proximations of a set (i.e., boundary region) provides a ba-
Rough sets introduced by Pawlak in 1981 (77, 78) and elab- sis for the roughness of an approximation. This is im-
orated in (13,33,34,67,68,74,7981) offer another approach portant because vagueness is allocated to some regions of
to CI by drawing attention to the importance of set approx- what is known as the universe of discourse (space) rather
imation in knowledge discovery and information granula- than to the whole space as encountered in fuzzy sets. The
 Computational Intelligence 3

study of what it means to be a part of provides a basis for sue. Computation Intelligence: An International Journal, vol.
what is known as mereology introduced by Lesniewski in 17, no. 3, 2001, 399603.
1927 (36). More recently, the study of what it means to be 7. A. Skowron, S. K. Pal (Eds.), Rough Sets, Pattern Recognition
a part of to a degree has led to a calculus of granules (8,37 and Data Mining Special Issue. Pattern Recognition Letters,
39,71,73). In effect, granular computing allows us to quan- vol. 24, no. 6, 2003, 829933.
tify uncertainty and take advantage of uncertainty rather 8. A. Skowron, Toward intelligent systems: Calculi of informa-
than blindly discarding it. tion granules. In: T. Terano, T. Nishida, A. Namatane, S.
Approximation spaces introduced by Pawlak (77), elab- Tsumoto, Y. Ohsawa, T. Washio (Eds.), New Frontiers in Arti-
orated by (33,34,66,69,7073), applied in (68,40,46,59,64) cial Intelligence, Lecture Notes in Articial Intelligence 2253.
Berlin: Springer-Verlag, 2001, 2839.
serve as a formal counterpart of our perception ability or
observation (69), and provide a framework for approximate 9. J. F. Peters, A. Skowron, J. Stepaniuk, S. Ramanna, Towards
an ontology of approximate reason, Fundamenta Informaticae,
reasoning about vague concepts. In its simplest form, an
vol. 51, nos. 12, 2002, 157173.
approximation space is any pair (U, R), where U is a non-
10. R. Marks, Intelligence: Computational versus Articial, IEEE
empty set of objects (called a universe of discourse) and R
Trans. on Neural Networks, 4, 1993, 737739.
is an equivalence relation on U (called an indiscernibililty
11. D. Fogel, Review of Computational Intelligence: Imitating
relation). Equivalence classes of an indiscernibility rela-
Life, IEEE Trans. on Neural Networks, 6, 1995, 15621565.
tion are called elementary sets (or information granules)
12. J. F. Peters, Time and Clock Information Systems: Concepts
determined by R. Given an approximation space S = (U,
and Roughly Fuzzy Petri Net Models. In: J. Kacprzyk (Ed.),
R), a subset X of U is denable if it can be represented as Knowledge Discovery and Rough Sets. Berlin: Physica Verlag,
the union of some of the elementary sets determined by a division of Springer Verlag, 1998.
R. It was originally observed that not all subsets of U are 13. Z. Pawlak, A. Skowron, Rudiments of rough sets, Information
denable in S (69). Given a non-denable subset X of U, Sciences, 177, 2006, 327. See, also, J. F. Peters, A. Skowron,
our observation restricted by R causes X to be perceived as Zdzislaw Pawlak life and work (19262006), Information Sci-
a vague object. An upper approximation B*X is the least ences, 177, 12, Z. Pawlak, A. Skowron, Rough sets: Some ex-
denable subset of U containing X, and the lower approxi- tensions, Information Sciences, 177, 2840 and Z. Pawlak, A.
mation B X is the greatest denable subset of U contained Skowron, Rough sets and Boolean reasoning, Information Sci-
in X. ences, 177, 4173.
Fuzzy set theory and rough set theory taken singly and 14. J. H. Holland, Adaptive plans optimal for payoff-only environ-
in combination pave the way for a variety of approximate ments, Proc. of the Second Hawaii Int. Conf. on System Sci-
reasoning systems and applications representing a syn- ences, 1969, 917920.
ergy of technologies from computational intelligence. This 15. J. R. Koza, Genetic Programming: On the Progamming of Com-
synergy can be found, for example, in recent work on the re- puters by Means of Natural Selection. Cambridge, MA: MIT
lation between fuzzy sets and rough sets (13,35,46,60,65), Press, 1993.
rough mereology (3739,65,66), rough control (42, 43), 16. C. Darwin, On the Origin of Species by Means of Natural Se-
fuzzy-rough-evolutionary control (44), machine learning lection, or the Preservation of Favoured Races in the Struggle
for Life. London: John Murray, 1959.
(34,45,59), fuzzy neurocomputing (3), rough neurocomput-
ing (46) diagnostic systems (34, 47), multi-agent systems 17. L. Chambers, Practical Handbook of Genetic Algorithms, vol.
I. Boca Raton, FL: CRC Press, 1995.
(8,9,48), real-time decision-making (12, 49), robotics and
unmanned vehicles (5053), signal analysis (55), and soft- 18. L. J. Fogel, A. J. Owens, M. J. Walsh, Articial Intelligence
through Simulated Evolution, Chichester, J. Wiley, 1966.
ware engineering (4,5558).
19. L. J. Fogel, On the organization of the intellect. Ph.D. diss.,
UCLA, 1964.
20. R. R. Yager and D. P. Filev, Essentials of Fuzzy Modeling and
BIBLIOGRAPHY Control. NY: John Wiley & Sons, Inc., 1994.
21. L. A. Zadeh, Fuzzy sets, Information and Control, 8, 1965,
1. J. C. Bezdek, On the relationship between neural networks, 338353.
pattern recognition and intelligence, Int. J. Approximate Rea- 22. L. A. Zadeh, Outline of a new approach to the analysis of com-
soning, 6, 1992, 85107. plex systems and decision processes, IEEE Trans. on Systems,
2. J. C. Bezdek, What is Computational Intelligence? In: J. Zu- Man, and Cybernetics, 2, 1973, 2844.
rada, R. Marks, C. Robinson (Eds.), Computational Intelli- 23. W. Pedrycz, Fuzzy Control and Fuzzy Systems, NY: John Wiley
gence: Imitating Life, Piscataway, IEEE Press, 1994, 112. & Sons, Inc., 1993.
3. W. Pedrycz, Computational Intelligence: An Introduction. Boca 24. R. Kruse, J. Gebhardt, F. Klawonn, Foundations of Fuzzy Sys-
Raton, FL: CRC Press, 1998. tems. NY: John Wiley & Sons, Inc., 1994.
4. W. Pedrycz, J. F. Peters (Eds.), Computational Intelligence 25. W. S. McCulloch, W. Pitts, A logical calculus of ideas imma-
in Software Engineering, Advances in Fuzzy Systems nent in nervous activity, Bulletin of Mathematical Biophysics
Applications and Theory, vol. 16. Singapore: World Scientic, 5, 1943, 115133.
1998. 26. F. Rosenblatt, Principles of Neurodynamics: Perceptrons and
5. D. Poole, A. Mackworth, R. Goebel, Computational Intelli- the Theory of Brain Mechanisms, Washington: Spartan Press,
gence: A Logical Approach. Oxford: Oxford University Press, 1961.
1998. 27. M. Minsky, S. Pappert, Perceptrons: An Introduction to Com-
6. N. Cercone, A. Skowron, N. Zhong (Eds.), Rough Sets, Fuzzy putational Geometry, Cambridge: MIT Press, 1969.
Sets, Data Mining, and Granular-Soft Computing Special Is-
4 Computational Intelligence

28. E. Fiesler, R. Beale (Eds.), Handbook on Neural Computation. Sets and Data Mining: Analysis for Imprecise Data. Norwell,
UK: Institute of Physics Publishing and Oxford University MA, Kluwer Academic Publishers, 1997, 91108.
Press, 1997. 46. S. K. Pal, L. Polkowski, A. Skowron (Eds.), Rough-Neuro
29. C. M. Bishop, Neural Networks for Pattern Recognition. Ox- Computing: Techniques for Computing with Words. Berlin:
ford: Oxford University Press, 1995. Springer-Verlag, 2003.
30. B. Widrow, M. E. Hoff, Adaptive switching circuits, Proc. IRE 47. R. Hashemi, B. Pearce, R. Arani, W. Hinson, M. Paule, A fu-
WESCON Convention Record, Part 4, 1960, 96104. sion of rough sets, modied rough sets, and genetic algorithms
31. B. Widrow, Generalization and information storage in net- for hybrid diagnostic systems. In: T. Y. Lin, N. Cercone (Eds.),
works of adaline neurons. In M. C. Yovits, G. T. Jacobi, G. D. Rough Sets and Data Mining: Analysis for Imprecise Data.
Goldstein (Eds.), Self-Organizing Systems. Washington, Spar- Norwell, MA, Kluwer Academic Publishers, 1997, 149176.
tan, 1962. 48. R. Ras, Resolving queries through cooperation in multi-agent
32. J. A. Freeman and D. M. Skapura, Neural Networks: Algo- systems. InT. Y. Lin, N. Cercone (Eds.), Rough Sets and Data
rithms, Applications and Programming Techniques. Reading, Mining: Analysis for Imprecise Data. Norwell, MA, Kluwer
MA, Addison-Wesley, 1991. Academic Publishers, 1997, 239258.
33. Z. Pawlak, Rough sets, Int. J. of Information and Computer 49. A. Skowron, Z. Suraj, A parallel algorithm for real-time deci-
Sciences, vol. 11, no. 5, 1982, 341356, 1982 sion making: a rough set approach. J. of Intelligent Systems,
34. Z. Pawlak, Rough Sets. Theoretical Aspects of Reasoning about vol. 7, 1996, 528.
Data, Dordrecht, Kluwer Academic Publishers, 1991. 50. J. F. Peters, T. C. Ahn, M. Borkowski, V. Degtyaryov, S. Ra-
35. W. Pedrycz, Shadowed sets: Representing and processing fuzzy manna, Line-crawling robot navigation: A rough neurocom-
sets, IEEE Trans. on Systems, Man, and Cybernetics, Part B: puting approach. In:C. Zhou, D. Maravall, D. Ruan (Eds.),
Cybernetics, 28/1, Feb. 1998, 103108. Autonomous Robotic Systems. Berlin: Physica-Verlag, 2003,
141164.
36. S. Lesniewski, O podstawach matematyki (in Polish), Przeglad
Filozoczny, vol. 30, 164206, vol. 31, 261291, vol. 32, 60101, 51. J. F. Peters, T. C. Ahn, M. Borkowski, Object-classication by a
and vol. 33, 142170, 1927. line-crawling robot: A rough neurocomputing approach. In:J.
J. Alpigini, J. F. Peters, A. Skowron, N. Zhong (Eds.), Rough
37. L. Polkowski and A. Skowron, Implementing fuzzy contain-
Sets and Current Trends in Computing, LNAI 2475. Springer-
ment via rough inclusions: Rough mereological approach
Verlag, Berlin, 2002, 595601.
to distributed problem solving, Proc. Fifth IEEE Int. Conf.
on Fuzzy Systems, vol. 2, New Orleans, Sept. 811, 1996, 52. M. S. Szczuka, N. H. Son, Analysis of image sequences for un-
11471153. manned aerial vehicles. In:M. Inuiguchi, S. Hirano, S. Tsumoto
(Eds.), Rough Set Theory and Granular Computing. Berlin:
38. L. Polkowski, A. Skowron, Rough mereology: A new paradigm
Springer-Verlag, 2003, 291300.
for approximate reasoning, International Journal of Approxi-
mate Reasoning, vol. 15, no. 4, 1996, 333365. 53. H. S. Son, A. Skowron, M. Szczuka, Situation identication by
unmanned aerial vehicle. In: Proc. of CS&P 2000, Informatik
39. L. Polkowski, A. Skowron, Rough mereological calculi of gran-
Berichte, Humboldt-Universitat zu Berlin, 2000, 177188.
ules: A rough set approach to computation, Computational
Intelligence: An International Journal, vol. 17, no. 3, 2001, 54. J. F. Peters, L. Han, S. Ramanna, Rough neural computing in
472492. signal analysis, Computational Intelligence, vol. 1, no. 3, 2001,
493513.
40. Bazan, H. S. Nguyen, A. Skowron, M. Szczuka: A view on rough
set concept approximation, In: G. Wang, Q. Liu, Y. Y. Yao, A. 55. J. F. Peters, S. Ramanna, Towards a software change classica-
Skowron, Proceedings of the Ninth International Conference tion system: A rough set approach, Software Quality Journal,
on Rough Sets, Fuzzy Sets, Data Mining and Granular Com- vol. 11, no. 2, 2003, 87120.
puting RSFDGrC2003), Chongqing, China, 2003, LNAI 2639, 56. M. Reformat, W. Pedrycz, N. J. Pizzi, Software quality analysis
181188. with the use of computational intelligence, Information and
41. J. Bazan, H. S. Nguyen, J. F. Peters, A. Skowron, M. Szczuka: Software Technology, 45, 2003, 405417.
Rough set approach to pattern extraction from classiers, Pro- 57. J. F. Peters, S. Ramanna, A rough sets approach to assess-
ceedings of the Workshop on Rough Sets in Knowledge Dis- ing software quality: Concepts and rough Petri net models.
covery and Soft Computing at ETAPS2003, April 1213, 23, In:S. K. Pal andA. Skowron (Eds.), Rough-Fuzzy Hybridiza-
Warsaw University, electronic version in Electronic Notes in tion: New Trends in Decision Making. Berlin: Springer-Verlag,
Computer Science, Elsevier, 2029. G. Frege, Grundlagen der 1999, 349380.
Arithmetik, 2, Verlag von Herman Pohle, Jena, 1893. 58. W. Pedrycz, L. Han, J. F. Peters, S. Ramanna, R. Zhai, Cali-
42. T. Munakata, Z. Pawlak, Rough control: Application of rough bration of software quality: Fuzzy neural and rough neural
set theory to control, Proc. Eur. Congr. Fuzzy Intell. Technol. approaches. Neurocomputing, vol. 36, 2001, 149170.
EUFIT96, 1996, 209218. 59. J. F. Peters, C. Henry, Reinforcement learning with approx-
43. J. F. Peters, A. Skowron, Z. Suraj, An application of rough set imation spaces. Fundamenta Informaticae, 71 (2-3), 2006,
methods to automatic concurrent control design, Fundamenta 323349.
Informaticae, 43(14), 2000, 269290. 60. W. Pedrycz, Granular computing with shadowed sets. In:D.
44. T. Y. Lin, Fuzzy controllers: An integrated approach based on Slezak, G. Wang, M. Szczuka, I. Duntsch, Y. Yao (Eds.),Rough
fuzzy logic, rough sets, and evolutionary computing. In: T. Y. Sets, Fuzzy Sets, Data Mining, and Granular Computing,
Lin, N. Cercone (Eds.), Rough Sets and Data Mining: Analysis LNAI 3641. Springer, Berlin 2005, 2331.
for Imprecise Data. Norwell, MA, Kluwer Academic Publish- 61. W. Pedryz, Granular computing in knowledge integration and
ers, 1997, 109122. reuse. In:D. Zhang, T. M. Khoshgoftaar, M.-L. Shyu (Eds.),
45. J. Grzymala-Busse, S. Y. Sedelow, W. A. Sedelow, Machine IEEE Int. Conf. on Information Reuse and Integration. Las
learning & knowledge acquisition, rough sets, and the En- Vegas, NV, USA, 1517 Aug. 2005.
glish semantic code. In: T. Y. Lin, N. Cercone (Eds.), Rough
 Computational Intelligence 5

62. W. Pedrycz, G. Succi, Genetic granular classiers in modeling JAMES F. PETERS

software quality. Journal of Systems and Software, 76(3), 2005, WITOLD PEDRYCZ
277285. University of Manitoba,
63. W. Pedrycz, M. Reformat, Genetically optimized logic models. Manitoba, Canada
Fuzzy Sets & Systems, 150(2), 2005, 351371. University of Alberta, Alberta,
64. J. F. Peters, Rough ethology: Towards a biologically-inspired Canada
study of collective behavior in intelligent systems with approx-
imation spaces. Transactions on Rough Sets III, LNCS 3400,
2005, 153174.
65. L. Polkowski, Rough mereology as a link between rough and
fuzzy set theories: A survey. Transactions on Rough Sets II,
LNCS 3135, 2004, 253277.
66. L. Polkowski, Rough Sets. Mathematical Foundations. Ad-
vances in Soft Computing, Physica-Verlag, Heidelberg, 2002.

67. Z. Pawlak, Some issues on rough sets. Transactions on Rough

Sets I, LNCS 3100, 2004, 158.
68. Z. Pawlak, A treatise on rough sets. Transactions on Rough
Sets IV, LNCS 3700, 2005, 117.
69. E. Orlowska, Semantics of Vague Concepts. Applications of
Rough Sets. Institute for Computer Science, Polish Academy
of Sciences, Report 469, March 1981.
70. A. Skowron, J. Stepaniuk, Generalized approximation spaces.
In:Lin, T. Y., Wildberger, A. M. (Eds.), Soft Computing, Simu-
lation Councils, San Diego, 1995, 1821.
71. A. Skowron, J. Stepaniuk, J. F. Peters, R. Swiniarski, Calculi
of approximation spaces. Fundamenta Informaticae, 72 (13),
2006, 363378.
72. A. Skowron, J. Stepaniuk, Tolerance approximation spaces.
Fundamenta Informaticae, 27(23), 1996, 245253.
73. A. Skowron, R. Swiniarski, P. Synak, P., Approximation spaces
and information granulation. Transactions on Rough Sets III,
LNCS 3400, 2005, 175189.
74. A. Skowron, J. F. Peters, Rough sets: Trends and Challenges.
In:Wang, G.Liu, Q., Yao, Y., Skowron, A. (Eds.), Proceedings
9th Int. Conf. on Rough Sets, Fuzzy Sets, Data Mining, and
Granular Computing (RSFDGrC2003), LNAI 2639, Springer-
Verlag, Berlin, 2003, 2534.
75. IEEE World Congress on Computational Intelligence. Vancou-
ver, B.C., Canada, 1621 July 2006.
76. Hamza, M. H. (Ed.), Proceedings of the IASTED Int. Conf.
on Computational Intelligence. Calgary, AB, Canada, 46 July
2005.
77. Z. Pawlak, Classication of Objects by Means of Attributes.
Institute for Computer Science, Polish Academy of Sciences,
Report 429, March 1981.
78. Z. Pawlak,Rough Sets. Institute for Computer Science, Polish
Academy of Sciences, Report 431, 1981.
79. Z. Pawlak, Rough classication. Int. J. of Man-Machine Stud-
ies, 20 (5), 1984, 127134.
80. Z. Pawlak, Rough sets and intelligent data analysis, Informa-
tion Sciences: An International Journal, 147 (14), 2002, 112.

81. Z. Pawlak, Rough sets, decision algorithms and Bayes theo-

rem, European Journal of Operational Research, 136, 2002,
181189.
82. E. P. Kement, R. Mesiar, E. Pap, Triangular Norms, Kluwer,
Dordrecht, 2000.
83. J. Bazan, A. Skowron, R. Swiniarski, Rough sets and vague
concept approximation: From sample approximation to adap-
tive learning, Transactions on Rough Sets V, LNCS 4100,
Springer, Heidelberg, 2006, 3962.