Evolving Intelligent Systems: A Comprehensive Analysis

Desklib provides solved assignments and past papers to help students succeed.

Contents
Introduction................................................................................................................................2
Developing of Intelligent Systems.............................................................................................2
System of Evolving Connectionist.............................................................................................2
Framework of ECOS..................................................................................................................4
Evolving of Fuzzy Systems....................................................................................................5
eTS Learning Between Clustering and Least Square.............................................................5
Mamdani Model.....................................................................................................................5
Applications...............................................................................................................................5
Conclusion..................................................................................................................................6
References..................................................................................................................................7
List of Figures
Figure 1: Fuzzy architecture.......................................................................................................3
Figure 2: Framework of eCI.......................................................................................................4

Introduction
The assignment describes Evolving of intelligent systems. The assignment includes different
applications and methods of an intelligent system. This system uses fuzzy neural networks
which are of different type like dynamic and neural fuzzy system. The dynamic method
includes unsupervised clustering technique while neural fuzzy system includes supervised
clustering technique. An intelligent system is used everywhere and there are many
applications of this system. This system is used in bioinformatics, medical and industrial.
Uses of intelligent systems are increases day by day. This method uses machine learning
which helps in providing better results on the basis of previous result.
Developing of Intelligent Systems
It is a system that is used for reasoning and decision making and helps in increasing machine
learning quotient of system which contains fuzzy logic, machine learning, and neural
networks. An intelligent system is used in one-line mode which has real-time applications in
industry, biology and in defense. The system parameter is based on evolving model structure
which is capable of adjusting any changes in object.
System of Evolving Connectionist
This system is categorized into different parts which include general principles, fuzzy-
networks, and neural networks. Here is the description of all the methods which are used in
ECOS [2].
 Principle of ECOS: ECOS is based on the connectionist system which is used to
change functionality and structure in on-line, continuous for incoming information.
They can easily process knowledge and data in both supervised and unsupervised
way. Mainly ECOS learn models from data by using clustering and uses output
function for every cluster. These clusters are created on the basis of data samples in
input and output space. Clusters are adjusted on the basis of new samples of data and
new clusters created. The similarity between rule node N=(W1,W2) and sample
S=(x,y) is measured using different ways and the most widely used technique is
Euclidean approach. This is most important approach which provides more accurate
results as compared to the other different approaches. ECOS is a method in which it
learns from data and information and then creates and update output function for each
of the cluster. Function is represented in W2 weight of connection which helps in

creating local models very easily. In this system each model is presented as local rule
which contains some information like area of cluster and an output function that is
applied to data in cluster.
 Fuzzy Neural Networks: The architecture of fuzzy neural contains 5 forward layers
and a layer of feedback for neurons. This architecture is divided into many different
fuzzy layers and each layer contains some information. It contains input, output and
one layer of fuzzy for input and output, a rule layer is also included. The figure
describes different layers of fuzzy architecture. The third layer includes rule nodes
which are evolved by learning (supervised and unsupervised). This rule node is used
for describing prototypes of output and input data. Rule node describes two vectors
through weight W1 and W2 and later this is adjusted by supervised learning on errors
of output. The fourth layer includes representation quantization of fuzzy for variable
of output. The fifth layer is used to represent values for productivity variables. The
evolving process is based on two assumptions, the first assumptions include that rule
nodes are prior to learning and connections are provided during learning and second
assumption is that all nodes are shaped in between evolving process. Rule node
presents an association between hyper-sphere from fuzzy input and output. Using
associate rule, a new data is vector to rule node. Then this node is adjusted in input
space of fuzzy but it depends on learning rate of r1 and output space is be contingent
on learning rate of r2. Change of the center to its new position is presented in
mathematics by changing weight rules from w1 to w2 [4].
Figure 1: Fuzzy architecture
 Inference of Neuro-fuzzy: EFuNN is depended on the neural networks of fuzzy
which is used to evolve functionality and structure using different types of clustering

techniques. But mainly supervised clustering technique is used. This method is based
on dynamic inference of fuzzy system which forms Takagi fuzzy rules to put up data
in unsupervised form. The main difference between fuzzy neural and dynamic fuzzy
neural system is that neural fuzzy system uses supervised clustering method. While
dynamic method uses unsupervised form of clustering. In dynamic method new rules
are formed and updated in the operation of system. To calculate the output in
dynamic method m-most rules are selected from previous rules of fuzzy set. There
are following benefits of ECOS model. These benefits of ECOS describes how it is
very important in the intelligent analytics system and helps to predict by previous
data. It is used in many fields which include industrial, medical and bioinformatics.
The benefits include fast learning, an adaptation of on-line and uses open structure. It
also includes insertion of rules and extraction. Integrate knowledge and data models
[1].
Framework of ECOS
The framework of eCI needs different evolving models. This framework includes the
decision part of high level, many e-models, and adaptation part. In this framework
environment is interact with extraction part.
Figure 2: Framework of eCI

Evolving of Fuzzy Systems
Fuzzy systems use some rules with generic consequents and antecedents. The fuzzy model
includes many structures that depend on consequents [3].
It is the first type of Takagi in which consequents are equal to linear functions. Another
model is of Mamdani model in which consequents are changed by defuzzified value. These
consequent models can be of any nonlinear function type which consists of model of
nonlinear and mapping of neural.
eTS Learning Between Clustering and Least Square
EFS is achieved when supervised is combined with unsupervised learning. Each rule of fuzzy
is to operate in sub-parts of input or output data spaces. eTS cluster input and output spaces to
N levels of fuzzy regions. eTS learning is used to expand subtractive into real-time by
transforming it into a strong tool for the use of on-line learning. This learning method is
created on calculation which is related to recursive of potential.
Mamdani Model
This sM model is a special case of the TS model. fSPC is an approach that helps in
developing EFS that are created on sM model. fSPC is an algorithm which is stimulated by
Statistical process, it is a method which is used to monitoring the variability. SPC method
cluster outputs of system into other cluster and relate them with same process model
behaviour. Boundaries of output system is define by hotelling statistics. The output which are
satisfying the condition is updated by the smoothing process.
Applications
EFS is used in a wide range of applications which includes industrial, medical and
bioinformatics and many more. Here are some applications of Intelligent system analytics [5].
 Industrial Application: A simple EFS model is useful to predict the thickness of paint
film on vehicle body which includes certain combinations of factors. Mainly these
variables consist of two things: 1 flow rate of fluid, it directly affects film thickness
and second is air draft velocity it includes humidity and temperature. The air
provided to booth to eliminate paint from spray. EFS provides a relationship between
thickness of paint film and variables on vertical and horizontal surfaces. Due to this
vehicle body contains three sM models: horizontal surface and left, right verticals.

X*(i), y*(i), s*(i) these are three models that cover horizontal surface. The
information related to these models is shown in figure.
 Bioinformatics: This is the method that consists of data that is related to biological
storage, modeling, analysis, and knowledge of discovery. There are many problems
in bioinformatics which are solved by using eCI. Here are some sub-parts of
bioinformatics which describes the use and application of intelligent analytics.
(a) Data analysis of micro-array gene: There are 5 rules of EFuNN which are used
for representing samples of clustering using a good prognosis class and bad
prognosis class.
(b) Modeling of Gene regulatory: This method is used for showing an interaction
between different gene cells. There are 2 methods that are used to derive GRN it
includes DENFIS and EFuNN. Mainly this model is used to predict gene future
values.
 Business forecast and financial: To predict future values of the business and financial
eci is used for finding future results on the basis of existing data. Below is a figure
which describes one-line prediction of week which shows exchange of rate.
 Brain Study: This method is also used in brain models which are used to collect the
data from the brain using the ECG. Below is a figure describes brain study.

Conclusion
The assignment provides a detailed description of intelligent system which includes different
methods, applications, and learnings. The system is working on the fuzzy neural networks
and architecture contains different layers and rule nodes. But there many optional layers are
present in architecture. It contains an input layer and output layers and some other layers
which are lies between input and output layer. There are many applications and benefits of
using intelligent system. This method uses many different types of techniques and methods
which helps in finding better results.

References
[1] de Jesús Rubio, J., USNFIS: uniform stable neuro-fuzzy inference
system. Neurocomputing, 262, 2017,pp.57-66.
[2] Pratama, M., Lu, J. and Zhang, G., Evolving type-2 fuzzy classifier. IEEE Transactions
on Fuzzy Systems, 24(3), 2015, pp.574-589.
[3] Stanley, K.O., Clune, J., Lehman, J. and Miikkulainen, R., Designing neural networks
through neuroevolution. Nature Machine Intelligence, 1(1), 2019, pp.24-35.
[4] Martinel, N., Micheloni, C., and Foresti, G.L., The evolution of neural learning systems: a
novel architecture combining the strengths of NTs, CNNs, and ELMs. IEEE Systems, Man,
and Cybernetics Magazine, 1(3), 2015,pp.17-26.
[5] de Campos, L.M.L., de Oliveira, R.C.L. and Roisenberg, M., Optimization of neural
networks through grammatical evolution and a genetic algorithm. Expert Systems with
Applications, 56, 2016,pp.368-384.