1
Computational Intelligence Theory: An Orientation Technique

S. Jaisiva1*, C. Kumar2, S. Sakthiya Ram3, C. Sakthi Gokul Rajan1 and P. Praveen Kumar4

1Electrical and Electronics Engineering, Sri Krishna College of Technology, Coimbatore, Tamil Nadu, India

2Electrical and Electronics Engineering, Karpagam College of Engineering, Coimbatore, Tamil Nadu, India

3Electronics and Instrumentation Engineering, Bannari Amman Institute of Technology, Sathyamangalam, Tamil Nadu, India

4Department of Information Technology, Sri Manakula Vinayagar Engineering College, Pondicherry, India

Abstract

The ability of a system to change its behavior to reach its objective in a variety of settings is intelligence. In reality, a different definition of computational intelligence (CI) is that it entails real-world adaption in challenging and shifting situations. In other words, it serves as a precise illustration of a notion. Adaptation and computational intelligence are intimately linked concepts. The concept, design, implementation, and advancement of computing paradigms driven by natural and cognitive motivations is known as CI. Evolutionary computation, fuzzy systems, and neural networks have historically been the three major foundations of CI. However, over time, various computing models that were inspired by nature have emerged. Sustainable smart information system, such as the creation of video games and cognitive developmental systems, heavily relies on CI. Deep learning study, especially that on deep convolutional neural networks, has exploded in recent years. Deep learning is currently the main approach for artificial intelligence. Deep learning has become the main technology for AI. In reality, CI is the foundation of some of the most effective AI systems.

Keywords: Computational intelligence, artificial intelligence, biological intelligence, neural networks, fuzzy systems, optimization, evolutionary computation

1.1 Computational Intelligence

Intelligence is a trait shared by all decision-makers with a goal. An analysis paradigm known as an artificial neural network (ANN) is loosely framed on the basis of the human brain massively parallel architecture [1]. It replicates a massively parallel, linked computing framework with a large number of very straightforward individual processing components (PEs). The phrases artificial neural network and neural network will now be used equally throughout this chapter. Fuzzies are non-statistical inexactitude and ambiguity in info, as used in this article. The majority of notions used or expressed in the real world are hazy. For instance, the sentence “It’s somewhat misty outdoors right now” combines the notions of being pretty and, even before, a long period of time. (One may even contend that the term is ambiguous and inaccurate enough to be hazy.) Fuzzy sets simulate the characteristics of estimation, ambiguity, and inaccuracy. Fuzzy membership values in a fuzzy set represent the membership dimensions (or grades) of the set’s components. It will be demonstrated that the fundamental concept of fuzzy set theory is a membership function, which is the same as a fuzzy set [2].

Crossover, mutations, and the survival of the fittest are examples of natural evolutionary phenomena that are incorporated into genetic algorithms, which are search algorithms. They are utilized for categorization as well as optimization more frequently. While genetic algorithms incorporate crossover, evolutionary programming approaches do not. Instead, they depend on mutation and the survival of the fittest. Comparable to genetic algorithms, evolution tactics frequently employ a distinct kind of mutation in addition to using combination to share data across members of population rather than crossover [3]. Computer programs can evolve using a technique called genetic programming. Hierarchical tree topologies are frequently used to manipulate structures. Potential solutions are dispersed throughout the problem space by particles in particle swarm optimization. The issue space’s chosen locations where prior fitness values have been high are where the particles are pushed. The term “computational intelligence” refers to a computing-based methodology that gives a system the capability to gain knowledge of novel situations, giving the system the appearance of possessing one or more rational qualities including generalization, discovery, connection, and abstraction. They are frequently made to resemble one or more characteristics of natural intelligence. In the illustration of a neural network paradigm is back-propagation, which presupposes a particular set of characteristics, such as the design and the learning algorithm [4]. A certain collection of options for each characteristic constitutes a paradigm. Introducing a separate paradigm includes putting together a group of characteristics that describe the desired behavior of the CI tool.

There are some words that should only be used with care. One such instance is neural networks, where it is important to be clear if we are taking about analytical tools for artificial neural network wetware. Let us explore the conceptual and technological underpinnings of computational intelligence tools and component approaches after providing the fundamental definitions [5]. We utilize and mention the caveat mentioned before. The creation of algorithmic models to address ever-more-complex issues is a key focus of algorithmic innovation. These clever algorithms are a subset of artificial intelligence, along with deductive reasoning, expert systems, case-based reasoning, and symbolic machine learning systems (AI). AI can be seen as a synthesis of various scientific areas, such as computer science, physiology, philosophy, sociology, and biology, just by looking at the broad range of AI methodologies [6].

Yet what exactly is intelligence? Definitions of intelligence continue to spark heated discussion. Dictionary definitions of intelligence include the capability for cognition and reason, as well as the capacity to perceive, comprehend, and benefit from experience (especially to a high degree). Innovation, ability, awareness, empathy, and instinct are other terms used to characterize characteristics of intelligence.

Can computers think for themselves? Even now, there is more disagreement over this issue than over how to define intelligence. Alan Turing gave this issue a lot of study in the middle of the 20th century. He thought it was possible to build devices that could duplicate the functions of the human brain. Turing firmly felt that a well-designed computer could perform every task that the brain was capable of. His predictions are still prophetic more than fifty years later. Smaller biological neural system components have been successfully modeled, but the complicated task of modeling is an essential component of mankind intelligence and remains unsolved [7].

The Turing test, created by Turing in 1950, is a measurement of computing intelligence. The test involved asking questions of both a person and a machine using a keyboard. The computer might be thought to be smart if the interviewer was unable to tell the computer from the person. Turing anticipated that by the year 2000, a system will be able to compete with the testing and training of 70% chance. Has his conviction been realized? In order to avoid jumping into yet another argument, the reader is left to choose the solution to this issue. However, the information in this book may help to clarify some aspects of the response [8].

The IEEE Neural Networks Council of 1996 gave a more modern version of artificial intelligence as the research of how to get computers to perform tasks that people are good at. These processes include the AI paradigms that can generalize, synthesize, discover, and connect as well as learn novel contexts. While specific approaches and techniques from various CI paradigms have been effectively used to address issues in the real world, the current trend is to create hybridization of models because no one model is always better than the others. By doing this, we strengthen the areas where each component of the hybrid CI system excels and do away with those where it falls short. Swarm intelligence is a category of the CI concepts, despite the fact that many investigators believe they should only fall within the category of synthetic biology [9].

1.2 Application Fields for Computational Intelligence

There are applications for which every computational intelligence element technique is particularly well suited. A particular problem might be solvable by either a neural network or a fuzzy system, but at varying standards of achievement; therefore, consider the fact that main applications may intersect. It might not even be typical of all the important application fields. It is intended to give some insight into the variety of issues that have been addressed by using CI’s component techniques.

1.2.1 Neural Networks

Generally speaking, neural networks are best suited for five types of applications. The first three have a connection.

1.2.1.1 Classification [10]

This section examines which of a number of predefined classes most accurately captures an input sequence. Usually, there are not many classes compared to the quantity of inputs. One illustration determines whether a specific EEG data section represents an epileptiform spike waveform. Another...