Computational

Intelligence
The expression computational intelligence (CI) usually refers to the
ability of a computer to learn a specific task from data or experimen-
tal observation. Even though it is commonly considered a synonym
of soft computing, there is still no commonly accepted definition of
computational intelligence.
The methods used are close to the human’s way of reasoning, i.e. it uses inexact and
incomplete knowledge, and it is able to produce control actions in an adaptive way. CI
therefore uses a combination of five main complementary techniques. The fuzzy logic
which enables the computer to understand natural language, artificial neural networks
which permits the system to learn experiential data by operating like the biological one,
evolutionary computing, which is based on the process of natural selection, learning the-
ory, and probabilistic methods which helps dealing with uncertainty imprecision.

Generally, computational intelligence is a set of nature-inspired computational methodologies and

approaches to address complex real-world problems to which mathematical or traditional modelling

can be useless for a few reasons: the processes might be too complex for mathematical reasoning,

it might contain some uncertainties during the process, or the process might simply be stochastic in

nature. Indeed, many real-life problems cannot be translated into binary language (unique values of

0 and 1) for computers to process it. Computational Intelligence therefore provides solutions for such

problems.

Except those main principles, currently popular approaches include biologically in-
spired algorithms such as swarm intelligence and artificial immune systems, which can
be seen as a part of evolutionary computation, image processing, data mining, natural
language processing, and artificial intelligence, which tends to be confused with Com-
putational Intelligence. But although both Computational Intelligence (CI) and Artificial
Intelligence (AI) seek similar goals, there’s a clear distinction between them.

Computational Intelligence is thus a way of performing like human beings. Indeed, the
characteristic of “intelligence” is usually attributed to humans. More recently, many
products and items also claim to be “intelligent”, an attribute which is directly linked to
the reasoning and decision making.

Computational intelligence
Contents
1 History
2 Difference between Computational and Artificial
Intelligence
3 The five main principles of CI and its applications
3.1 Fuzzy logic
3.2 Neural networks
3.3 Evolutionary computation
3.4 Learning theory
3.5 Probabilistic methods
4 See also
5 Notes
6 Publications
7 References
History
The notion of Computational Intelligence was first used by the IEEE Neural Networks
Council in 1990. This Council was originally founded in the 1980s by a group of research-
ers interested in the development of biological and artificial neural networks. On Novem-
ber 21, 2001, the IEEE Neural Networks Council became the IEEE Neural Networks Society,
to become the IEEE Computational Intelligence Society two years later by including new
areas of interest such as fuzzy systems and evolutionary computation, which they related
to Computational Intelligence in 2011 (Dote and Ovaska).

But the first clear definition of Computational Intelligence was introduced by Bezdek in
1994: a system is called computationally intelligent if it deals with low-level data such as
numerical data, has a pattern-recognition component and does not use knowledge in
the AI sense, and additionally when it begins to exhibit computational adaptively, fault
tolerance, speed approaching human-like turnaround and error rates that approximate
human performance.

Bezdek and Marks (1993) clearly differentiated CI from AI, by arguing that the first one is
based on soft computing methods, whereas AI is based on hard computing ones.

Computational intelligence
Difference between Computa-
tional and Artificial Intelligence
Although Artificial Intelligence and Computational Intelligence seek a similar long-term
goal: reach general intelligence, which is the intelligence of a machine that could per-
form any intellectual task that a human being can; there’s a clear difference between
them. According to Bezdek (1994), Computational Intelligence is a subset of Artificial
Intelligence.

Hard computing techniques work following binary logic based

There are two types of
on only two values (the Booleans true or false, 0 or 1) on which machine intelligence:
modern computers are based. One problem with this logic is the artificial one based
on hard computing
that our natural language cannot always be translated easi-
techniques and the
ly into absolute terms of 0 and 1. Soft computing techniques, computational one
based on fuzzy logic can be useful here. Much closer to the based on soft comput-
way the human brain works by aggregating data to partial ing methods, which
enable adaptation to
truths (Crisp/fuzzy systems), this logic is one of the main exclu- many situations.
sive aspects of CI.

Within the same principles of fuzzy and binary logics follow

crispy and fuzzy systems. Crisp logic is a part of artificial intelligence principles and con-
sists of either including an element in a set, or not, whereas fuzzy systems (CI) enable ele-
ments to be partially in a set. Following this logic, each element can be given a degree
of membership (from 0 to 1) and not exclusively one of these 2 values.
Fuzzy logic
As explained before, fuzzy logic, one of CI’s main principles, consists in measurements
and process modelling made for real life’s complex processes. It can face incomplete-
ness, and most importantly ignorance of data in a process model, contrarily to Artificial
Intelligence, which requires exact knowledge.

This technique tends to apply to a wide range of domains such as control, image pro-
cessing and decision making. But it is also well introduced in the field of household appli-
ances with washing machines, microwave ovens, etc. We can face it too when using a
video camera, where it helps stabilizing the image while holding the camera unsteadily.
Other areas such as medical diagnostics, foreign exchange trading and business strate-
gy selection are apart from this principle’s numbers of applications.

Fuzzy logic is mainly useful for approximate reasoning, and doesn’t have learning abili-
ties, a qualification much needed that human beings have. It enables them to improve
themselves by learning from their previous mistakes.

Computational intelligence
Neural networks
This is why CI experts work on the development of artificial neural networks based on
the biological ones, which can be defined by 3 main components: the cell-body which
processes the information, the axon, which is a device enabling the signal conducting,
and the synapse, which controls signals. Therefore, artificial neural networks are doted of
distributed information processing systems, enabling the process and the learning from
experiential data. Working like human beings, fault tolerance is also one of the main as-
sets of this principle.

The main ap- Concerning its applications, neural networks can be classified into five
plications of groups: data analysis and classification, associative memory, cluster-
Computational ing generation of patterns and control. Generally, this method aims to
Intelligence in-
analyze and classify medical data, proceed to face and fraud detec-
clude computer
science, engi- tion, and most importantly deal with nonlinearities of a system in order
neering, data to control it. Furthermore, neural networks techniques share with the
analysis and fuzzy logic ones the advantage of enabling data clustering.
bio-medicine.
Evolutionary computation
Based on the process of natural selection firstly introduced by Charles Robert
Darwin, the evolutionary computation consists in capitalizing on the strength of
natural evolution to bring up new artificial evolutionary methodologies.[11][page
needed] It also includes other areas such as evolution strategy, and evolutionary
algorithms which are seen as problem solvers... This principle’s main applications
cover areas such as optimization and multi-objective optimization, to which tradi-
tional mathematical one techniques aren’t enough anymore to apply to a wide
range of problems such as DNA Analysis, scheduling problems.

Learning theory
Still looking for a way of “reasoning” close to the humans’ one, learning theory is
one of the main approaches of CI. In psychology, learning is the process of bring-
ing together cognitive, emotional and environmental effects and experiences to
acquire, enhance or change knowledge, skills, values and world views (Ormrod,
1995; Illeris, 2004). Learning theories then helps understanding how these effects
and experiences are processed, and then helps making predictions based on
previous experience.

Computational intelligence
Probabilistic methods
Being one of the main elements of fuzzy logic, probabilistic methods firstly intro-
duced by Paul Erdos and Joel Spencer (1974), aim to evaluate the outcomes of a
Computation Intelligent system, mostly defined by randomness. Therefore, proba-
bilistic methods bring out the possible solutions to a reasoning problem, based on
prior knowledge.

Cognitive robotics
Cognitive robotics is concerned with endowing a robot with intelligent behavior
by providing it with a processing architecture that will allow it to learn and reason
about how to behave in response to complex goals in a complex world. Cogni-
tive robotics may be considered the engineering branch of embodied cognitive
science and embodied embedded cognition.

Core issues

While traditional cognitive modeling approaches have assumed symbolic coding

schemes as a means for depicting the world, translating the world into these kinds of
symbolic representations has proven to be problematic if not untenable. Perception and
action and the notion of symbolic representation are therefore core issues to be ad-
dressed in cognitive robotics.
 Starting point
Cognitive robotics views animal cognition as a starting point for the development
of robotic information processing, as opposed to more traditional Artificial Intelli-
gence techniques. Target robotic cognitive capabilities include perception pro-
cessing, attention allocation, anticipation, planning, complex motor coordination,
reasoning about other agents and perhaps even about their own mental states.
Robotic cognition embodies the behavior of intelligent agents in the physical
world (or a virtual world, in the case of simulated cognitive robotics). Ultimately
the robot must be able to act in the real world.

Learning techniques
Motor Babble

A preliminary robot learning technique called motor babbling involves correlating pseu-
do-random complex motor movements by the robot with resulting visual and/or auditory
feedback such that the robot may begin to expect a pattern of sensory feedback given
a pattern of motor output. Desired sensory feedback may then be used to inform a mo-
tor control signal. This is thought to be analogous to how a baby learns to reach for ob-
jects or learns to produce speech sounds. For simpler robot systems, where for instance
inverse kinematics may feasibly be used to transform anticipated feedback (desired
motor result) into motor output, this step may be skipped.

Computational intelligence
 Imitation

Once a robot can coordinate its motors to produce a desired result, the tech-
nique of learning by imitation may be used. The robot monitors the performance
of another agent and then the robot tries to imitate that agent. It is often a chal-
lenge to transform imitation information from a complex scene into a desired mo-
tor result for the robot. Note that imitation is a high-level form of cognitive behav-
ior and imitation is not necessarily required in a basic model of embodied animal
cognition.

Knowledge acquisition

A somewhat more directed mode of exploration can be A more com-

plex learning
achieved by “curiosity” algorithms, such as Intelligent
approach is “au-
Adaptive Curiosity or Category-Based Intrinsic Motivation. tonomous knowl-
These algorithms generally involve breaking sensory in- edge acquisi-
put into a finite number of categories and assigning some tion”: the robot is
sort of prediction system (such as an Artificial Neural Net- left to explore the
environment on
work) to each. The prediction system keeps track of the
its own. A system
error in its predictions over time. Reduction in prediction of goals and
error is considered learning. The robot then preferential- beliefs is typically
ly explores categories in which it is learning (or reducing assumed.
prediction error) the fastest.
 Other architectures

Some researchers in cognitive robotics have tried using architectures such as

(ACT-R and Soar (cognitive architecture)) as a basis of their cognitive robotics
programs. These highly modular symbol-processing architectures have been used
to simulate operator performance and human performance when modeling sim-
plistic and symbolized laboratory data. The idea is to extend these architectures
to handle real-world sensory input as that input continuously unfolds through time.
What is needed is a way to somehow translate the world into symbols.

Questions

Some of the fundamental questions to still be answered in cognitive robotics are:

- How much human programming should or can be involved to support

the learning processes?
- How can one quantify progress? Some of the adopted ways is the re-
ward and punishment. But what kind of reward and what kind of punishment?
In humans, when teaching a child for example, the reward would be candy or
some encouragement, and the punishment can take many forms. But what is an
effective way with robots?

Computational intelligence
 Books

Cognitive Robotics book by Hooman Samani, takes a multidiciplinary approach

to cover various aspects of cognitive robotics such as artificial intelligence, physi-
cal, chemical, philosophical, psychological, social, cultural, and ethical aspects