IEEE Computational Intelligence Magazine - November 2020 - 12

2) Fuzzy Logic
Fuzzy logic [10] can be seen as a formulation
defined by multi-valued logic. Meanwhile, the
true value of a variable's formulation can be
any real number between, but not limited to,
0 and 1. It is often utilized to solve the problem of uncertainty, where the truth value may
not be all true or all false. As a result, fuzzy
logic has been successfully applied in the field
of clinical realms, including a continuous
blood glucose prediction system and a tuberculosis diagnosis platform based on chest
X-ray, among other devices. We can also see this in use of a
video camera to help stabilize an image endoscope. Other
areas such as household appliances, business decision making,
and financial analysis are also examples of applications of this
principle [6]. A main application of fuzzy logic is approximate
reasoning. However, the methods of fuzzy logic reasoning usually lack learning abilities, which are necessary for a multitude
of tasks.

Based on our observations, there are several urgent
issues related to COVID-19 that must be combatted.
These issues can be categorized into five topics:
tracking and predicting virus propagation (TPVP),
characterization of symptoms of virus infections (CSVI),
treatment design (TrD), precaution development (PD),
and public health policy making (PHPM).
Compared to the hard-computing-based artificial intelligence,
computational intelligence can adapt to many different conditions via the benefits of the concept of soft-computing. Hard
computing techniques are designed using a Boolean logic
based only on true or false values that information engineering
relies on. One critical issue in Boolean logic is that Boolean
values are unable to interpret natural language easily. However,
based on fuzzy logic, soft computing techniques can deal with
uncertain cases. This type of logic is one proprietary aspect of
computational intelligence, and by aggregating data into partial
facts, it is approximated to the manner in which the human
brain acts [7].
B. Categorization of Computational Intelligence

As mentioned above, the notion of computational intelligence
has been around for 30 years. During this period, new concepts
have been constantly added to the field, thereby reinforcing the
discipline. Today, we can broadly divide computational intelligence techniques into five categories: neural networks, fuzzy
logic, evolutionary computation, computational learning theory, and probabilistic methods.
1) Neural Networks
Based on biological neural networks, an artificial neural network (called "neural network" for short) is designed as a network of artificial neurons or nodes. Artificial neural networks
can be used for regression or classification modeling for prediction and automatic control. A large number of simulation
data using limited data sets. This structure is the foundation of
deep learning, which is good at representation learning.
Accordingly, artificial neural networks process and learn information from data via the systems of distributed information
processing [8]. By doing so, one of the crucial properties of
artificial neural networks is fault tolerance, which is approximately modeled on the manner in which the human brain
operates [6]. Based on these characteristics, neural networks
have been widely applied to data analytics, clustering, classification, and automatic control engineering. In real-world
applications, such methods aim to analyze and classify medical
data, recognize human faces, detect computer fraud, and deal
with the nonlinearity of a system for better process control [9].
Furthermore, neural network techniques can incorporate
fuzzy logic concepts.

IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE | NOVEMBER 2020

3) Evolutionary Computation
Evolutionary computation (EC) is global optimization method
inspired by biological evolution [6]. It is a family of algorithms
and is a branch of computational intelligence and natural computing. EC systems solve problems via populations, error and
success, meta-heuristics, or stochastic optimization. An initial
set of candidate solutions is generated and updated iteratively,
such as the removal of less-desired solutions and the insertion
of noise. A population of solutions is subject to natural selection
or artificial selection and mutation, and therefore evolves and
adapts-i.e., increases fitness (function quantizes how adapted/
desired the solution is). EC is popular in computational intelligence because it results in near-optimal solutions in a wide
spectrum of contexts [11] where there are many variants and
extensions for specific data structures and problems.
4) Computational Learning Theory
Computational learning theory (referred to as learning theory
for short) is a sub-field of artificial intelligence mainly for the
research and development of learning strategies for machine
learning. Computational learning theory is one of the principal
methods in computational intelligence, which seeks for a way
to achieve reasoning that recapitulates human rezoning humans.
In psychology, learning is the process of enhancing or changing
knowledge, skills, values, and world views through cognition
and experience [6]. Inspired by psychology, computational
learning theory is utilized to actualize the process of experience
and decision making according to previous experiences.
5) Probabilistic Methods
A probabilistic method is a nonconstructive approach used to
prove the existence of specified types of mathematical objects.
It operates by showing that if one randomly selects an object
from a specified category, the probability that the result will

IEEE Computational Intelligence Magazine - November 2020

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - November 2020