ArticlePDF Available

Computer Aided Development of Fuzzy, Neural and Neuro-Fuzzy Systems

Authors:

Abstract and Figures

Development of an expert system is difficult because of two challenges involve in it. The first one is the expert system itself is high level system and deals with knowledge, which make is difficult to handle. Second, the systems development is more art and less science; hence there are little guidelines available about the development. This paper describes computer aided development of intelligent systems using modem artificial intelligence technology. The paper illustrates a design of a reusable generic framework to support friendly development of fuzzy, neural network and hybrid systems such as neuro-fuzzy system. The reusable component libraries for fuzzy logic based systems, neural network based system and hybrid system such as neuro-fuzzy system are developed and accommodated in this framework. The paper demonstrates code snippets, interface screens and class libraries overview with necessary technical details.
Content may be subject to copyright.
International Journal of Engineering and Applied Computer Science (IJEACS)
Volume: 02, Issue: 01, January 2017
ISBN: 978-0-9957075-2-8
www.ijeacs.com
10
Computer Aided Development of Fuzzy, Neural and
Neuro-Fuzzy Systems
Priti Srinivas Sajja
Department of Computer Science and Technology
Sardar Patel University
Vallabh Vidyanagar, India
Abstract Development of an expert system is difficult because of
two challenges involve in it. The first one is the expert system
itself is high level system and deals with knowledge, which make
is difficult to handle. Second, the systems development is more
art and less science; hence there are little guidelines available
about the development. This paper describes computer aided
development of intelligent systems using modem artificial
intelligence technology. The paper illustrates a design of a
reusable generic framework to support friendly development of
fuzzy, neural network and hybrid systems such as neuro-fuzzy
system. The reusable component libraries for fuzzy logic based
systems, neural network based system and hybrid system such as
neuro-fuzzy system are developed and accommodated in this
framework. The paper demonstrates code snippets, interface
screens and class libraries overview with necessary technical
details.
Keywords: Fuzzy logic, Neural network, Neuro-fuzzy systems, Soft
computing, Automatic development.
I. INTRODUCTION
Advancement of science and technology has been
increasingly utilized as a major tool for uplift of mankind.
Innovations of modern information and communication
technologies (ICT) made human life smoother and problem
solving has become easier. It is observed that usage of modern
ICT (like Internet) has become ambient and ubiquitous. In
spite of availability of lots of tools and technologies,
expectations from mighty machines are continuously
increasing and demand for more and more human like
intelligent systems in various fields has evolved.
Development of an intelligent system is a challenging job
due to several reasons. The prime among them are abstract
nature of knowledge, volume of knowledge, lack of knowledge
acquisition and representation techniques and lack of
models/quality standards for the development of an intelligent
system. Typical Artificial Intelligence (AI) techniques facilitate
development of intelligent system with the aforementioned
problems. However, they lack self learning, human like
interaction, processing and require lot of efforts as well as cost.
There are some new AI techniques such as bio-inspired
techniques that offer some advantages above the typical AI
techniques. These techniques include artificial neural network,
swarm intelligence, fuzzy logic etc. and their hybridization
(such as neuro-fuzzy approach). These techniques are
sometimes also known as soft computing techniques. Many
tools are available to develop such soft AI based intelligent
systems. However, these tools are costly, less user friendly and
application specific. Further, these tools need training and
practice to certify their efficient utilization. There is a
requirement of a generic tool which interacts through
native/natural language of non-computer professionals/users,
reduces effort of development of an intelligent system and
saves time of development. No such generic tool is available as
per the literature found at present. This paper presents a design
and implementation of generic tool that facilitates automatic
development of soft computing intelligent systems such as
neural network based systems, fuzzy logic based systems and
hybrid neuro-fuzzy systems in a given domain.
The paper organization is as follows. Section 2 of the paper
presents national and international scenario related to the
aforementioned problem and discusses limitations of the
existing solutions. The aims and objectives of the proposed
research work are enlisted in this section.
Section 3 describes architecture of the proposed framework
which can be utilized as generic and user friendly tool that
facilitates automatic development of soft computing intelligent
systems. An important component of the framework is library
of reusable components to develop intelligent systems. Design
and implementation issues of the centralized library along with
other necessary components are described here. At the end, the
paper concludes with advantages and application of the
framework generated in different other areas. The section also
presents future scope of the research work.
II. LITERATURE SURVEY
To develop soft computing system techniques such as
artificial neural network (ANN), type 1 and type 2 fuzzy logic
(T1FL and T2FL) and hybrid neuro-fuzzy systems are vastly
used. This section presents in brief about each techniques and
work done in different areas using the respective techniques.
Priti Srinivas Sajja
Volume: 02, Issue: 01, January 2017
ISBN: 978-0-9957075-2-8
www.ijeacs.com
11
A. Artificial Neural Network (ANN)
An Artificial Neural Network (ANN) is an interconnected
group of artificial neurons that uses a computational model for
processing information based on a connectionist approach to
computation to impart self learning and other characteristics
associated with intelligence. ANN is powerful tool for
modeling intelligent systems when self learning is most desired
feature and generalized knowledge is not available or can not
be documented easily. ANN has ability to identify and learn
the correlation between the input patterns, data or cases
provided to it. Due to this feature of ANN, it is increasingly
being used in intelligent systems where one has little or
incomplete understanding of the problem under research but
experience or training data of domain expert is readily
available. The popular models for implementing ANN are
Hopfield model, multilayer perceptron with supervised
learning, Kohenon model (self organizing map), and recurrent
network. Programming environment like java, c# and other
third generation programming languages help in implementing
ANN based systems. There are some packages like matlab
(www.mathworks.com/products/matlab) are also available
which facilitates development of such system.
B. Fuzzy Logic
Fuzzy logic is a multi-valued logic conceived by Zadeh [1]
and used to achieve human like loose categorization of objects
into classes without boundaries. The fuzzy logic is based on
fuzzy sets. Every given object has the partial (fuzzy)
belongingness of the concerned fuzzy sets, which is determined
by its membership function. For example, if the temperature is
18 degree centigrade, then it belongs fully (say 1) to the set of
comfortable temperature. If the temperature is 12 degree
centigrade, then it belongs partially (say 0.4) to the set of
comfortable temperature. Young lady, rich person, luxury car
and comfortable temperature are a few examples of linguistic
vague words which can be interpreted by machine with the
help of fuzzy membership functions, hence can be easily used
in ‘if…then…else’ rules. System with mainly such fuzzy
‘if…then…else’ rules are capable of representing knowledge
and are known as fuzzy rule based system. Mamdani [2] and
Takagi & Sugeno [3] are vastly utilized basic modeling
techniques for fuzzy logic based systems. Further, Jerry
Mendel has proposed an idea of type 2 fuzzy logic which was
an extension of the original fuzzy logic proposed by Prof.
Zadeh [1]. The basis of type 2 fuzzy logic is type 2 fuzzy sets.
Type 2 fuzzy sets incorporate uncertainty as extra third
dimension which gives much clear and logical information
about the problem under research. Further, type 2 fuzzy
systems need to accompany with type reducer component that
converts type 2 membership functions into simple fuzzy
membership functions. The popular tools to develop a fuzzy
logic based system are third generation programming
languages such as java and c# of .Net framework
(www.microsoft.com/net) and packages like Matlab
(www.mathworks.in). Many specific packages also have been
developed such as Fuzzy Attitude (www.fuzzytech.com),
JFuzzyLogic (sourceforge.net/projects/javafuzzyed/), Fuzzy
Editor (http://jeux.windows.simplenet.com/), etc.
C. Hybrid Systems
Every technique has its advantages and limitations. The
idea behind hybridization of two or more techniques is to get
advantages of all candidate technique in a common application.
For example, consider hybridization of ANN and FL. It is
observed that both ANN and FL have their own pros and cons.
ANN based systems are good where there is availability of data
but lack of generalized knowledge behind it. ANN systems are
good in self learning however lacks in documentation of
knowledge. This is the prime reason why such system cannot
provide reasoning and detail explanation of decision made.
Fuzzy logic is good in handling uncertainty and handling
natural linguistic values, but lacks self learning and enforces
documentation of knowledge in generalized form. Hybridizing
these two technologies provide dual advantages of FL and
ANN both in one common application. Specifically Neuro-
Fuzzy hybridization (NF) achieves advantages of self learning,
explanation and reasoning and user friendly interface along
with the advantages associated with documentation of
knowledge. As stated earlier, third generation programming
languages and packages like Matlab can be used to develop
such hybrid systems.
Fuzzy Adaptive Learning Control Network (FALCON) [4]
Adaptive Neuro Fuzzy Inference System (ANFIS) [5],
Generalized Approximated Reasoning based Intelligent Control
(GARIC) [6], Neuro-Fuzzy Control (NEFCON) [7], Fuzzy
Inference and Neural Network in Fuzzy Inference Software
(FINEST) [8], Fuzzy Net (FUN) [9], Evolving Fuzzy Neural
Network (EFuNN) [10], Self Constructing Neural Fuzzy
Inference Network (SONFIN) [11], etc are popular tools that
help in development of such hybrid system.
Not only NF system, but also different techniques such as
genetic algorithm and fuzzy logic, neuro-fuzzy-genetic and
neuro-genetic hybridization are also popular. At present our
framework was designed and implemented to support
development of FL, ANN and NF type of systems. However, it
can be enhanced to encompass other soft computing techniques
and tools. Hence, the literature survey is restricted to these
approaches only.
The above tools are either costly or application specific
such as Cihan H et al. [12]; or not web-enabled. Many of them
are not user friendly and to use such tools would be challenge
for the non-computer professionals. Tools like Adaptive
Network Fuzzy Inference System (ANFIS) and Dynamic
Evolving Neuro-Fuzzy Inference System (DENFIS) need
platform of MatLab which is costly. Many researchers have
experimented development of dedicated applications in the
field. Pioneer of them can be given as Ajit Abraham [13] Jang
et al. [14], Mendel [15] and Wu & Mendal [16], Emilio Soria-
Olivas et al. [17], Ching Long Su et al. [18], John & Coupland
[19], Oscar Castillo & Patricia Melin [20]. Much application
specific work is done by various researchers including
Malkawi & Murad [21], Nie et al. [22], Bouzaidaa et al. [23]
and Azriyenni & Mustafa [24].
Priti Srinivas Sajja
Volume: 02, Issue: 01, January 2017
ISBN: 978-0-9957075-2-8
www.ijeacs.com
12
Considering the aforementioned work; following
observations can be made.
Majority of the existing solutions are application
specific;
The solutions may not be web based;
The solution, which supports computer aided
development may not support modern artificial
intelligent techniques;
The existing solutions may not be flexible and
extendible to accommodate users new requirements
and other technology in future;
The exiting solutions/tools may not be reusable; and
The exiting solutions are generally meant for
programmers hence they may not be user-friendly to
non-computer professionals; etc.
This leads to a development of generic, web-enabled and
user friendly architecture that supports interactive development
of all type of soft computing intelligent system. As a prototype,
the framework is designed and implemented for automatic
development of FL, ANN and NF hybridized systems.
However, it is designed in a flexible way to support many more
latest technologies. The prime objectives were decided as
follows.
Development of fuzzy logic based editor to facilitate
working with linguistic variable and vague input;
Development of popular fuzzy membership functions
with fuzzification and defuzzification techniques;
Development of type reducer component for
converting type 2 fuzzy systems into typical fuzzy
systems so as the above fuzzy function can be used;
Figure 1. Multi layer framework.
Development of reusable library containing activation
functions and learning algorithms for artificial neural
network;
Mechanism for Neuro-fuzzy hybridization; and
Development of miscellaneous features such as back
up, security feature, web templates and theme library
of variety of presentation of generated code, etc;
Section 3 describes the design of the framework.
III. DESIGN OF THE FRAMEWORK
Design of the framework which meets objectives finalized
and mentioned in earlier section is presented here. The
framework is divided into three layers. In its first layer
repositories for reusable codes are stored as generic
independent objects. The reusable codes included here are
neural network, fuzzy logic and neuro-fuzzy systems. Second
layer is a database layer. The database accommodates third
party tool, user profiles, meta data repositories and protocols, if
any. Third layer is an interface layer. The interface layer
accommodates information acquisition and interaction facilities
for users of the system. This layer also provides facility of
customized representation of the output to the users of the
system. The interface layer also accommodates other
information such as local databases, log of the system and
frequently asked queries. Fig. 1 illustrates these layers in a
generic framework.
Using the framework one can generate artificial neural
network based systems, fuzzy logic based systems and neuro-
fuzzy systems at this stage. The framework is flexible enough
to add reusable component of other paradigms such as Genetic
Algorithms (GA). In this case, GA based systems can be
developed or neuro-genetic systems can be developed using the
framework.
A. Generation of ANN
The artificial neural network repository include code
segments for feed forward, radial basis function, Kohonen self
organized maps, learning vector quantization, recurrent
networks, etc. Some of the codes follow existing methodology
and some use novel mechanisms such as modified sigmoid and
tangent activation functions. When parameters of the ANN
such as number of layers and neurons, learning rate, learning
algorithm, etc; the ANN is generated. To generate any feed
forward, fully connected, back propagation type of multilayer
ANN with supervised learning following code snippet is used.
Repository Layer ………………. ……….. ………
ANN
FL
Database Layer
Third Party Tools
Meta Data
User Profiles
Interface Layer
Info Acquisition
Customized Presentation
Acquisition
Other Info
Profiles
Priti Srinivas Sajja
Volume: 02, Issue: 01, January 2017
ISBN: 978-0-9957075-2-8
www.ijeacs.com
13
public class NeuralNetwork
{ protected Layer[ ] layers;
protected int ni;
protected LearningAlgorithm la;
public int N_Inputs
{ get { return ni; } }
public int N_Outputs
{ { return layers[N_Layers - 1].N_Neurons; }
}
public int N_Layers
{ get { return layers.Length; } }
public LearningAlgorithm LearningAlg
{ { return la; }
set { la = (value != null) ? value : la; } }
public Layer this[int n]
{ { return layers[n]; }}
public NeuralNetwork(int inputs, int[] layers_desc,
ActivationFunction n_act, LearningAlgorithm learn)
{ if (layers_desc.Length < 1)
throw new Exception("PERCEPTRON : cannot build
perceptron, it must have at least 1 layer of neurone");
if (inputs < 1)
throw new Exception("PERCEPTRON : cannot
build perceptron, it must have at least 1 input");
la = learn;
ni = inputs;
layers = new Layer[layers_desc.Length];
layers[0] = new Layer(layers_desc[0], ni);
for (int i = 1; i < layers_desc.Length; i++)
layers[i] = new Layer(layers_desc[i], layers_desc[i
- 1], n_act);
}
public void randomizeWeight()
{ foreach (Layer l in layers)
l.randomizeWeight(); }
public void randomizeThreshold()
{ foreach (Layer l in layers)
l.randomizeThreshold(); }
public void randomizeAll()
{ foreach (Layer l in layers)
l.randomizeAll(); }
public void
setActivationFunction(ActivationFunction f)
{ foreach (Layer l in layers)
l.setActivationFunction(f); }
{ foreach (Layer l in layers)
l.setRandomizationInterval(min, max); }
public float[] Output(float[] input)
{ (input.Length != ni)
throw new Exception("PERCEPTRON : Wrong input
vector size, unable to compute output value");
float[] result;
result = layers[0].Output(input);
for (int i = 1; i < N_Layers; i++)
result = layers[i].Output(result);
return result;
}
} }
However, users are not aware of the background code; they
can interact through the framework by a friendly interface as
shown in Fig. 2.
B. Generation of Fuzzy Logic
To generate fuzzy logic based system the components such
as fuzzification method, defuzzification method, type reduction
codes (for type 2 FL systems), etc are developed and kept in
the repository layer. Many of these methods are innovative.
Fig. 3 provides an overview of the components of the FL
repositories.
Codes of each component shown in Fig. 3 are developed
and kept ready. To generate fuzzy logic based system, an
interface is generated as shown in Fig. 4.
Priti Srinivas Sajja
Volume: 02, Issue: 01, January 2017
ISBN: 978-0-9957075-2-8
www.ijeacs.com
14
Figure 2. Interface to create ANN.
Figure 3. FL repository components.
Fuzzy Logic
Membership Function
Linear, Triangular, Gaussian,
Trapezoidal, with other
modified functions, etc.
Defuzzification Methods
Centroid, Center of gravity,
max membership, Weighted
average, Mean max, etc.
Type Reducer
Centroid type, interval
integration type, heuristic
based type reduction, etc.
Knowledge Base
Rules, Temporary work
space, and Inference
mechanism etc.
Priti Srinivas Sajja
Volume: 02, Issue: 01, January 2017
ISBN: 978-0-9957075-2-8
www.ijeacs.com
15
Figure 4. Generation of FL based system.
Figure 5. Developing a hybrid neuro-fuzzy systems.
Priti Srinivas Sajja
Volume: 02, Issue: 01, January 2017
ISBN: 978-0-9957075-2-8
www.ijeacs.com
16
C. Generation of Neuro Fuzzy Component
Since the components of the ANN and FL are ready they
can be reused to develop hybrid neuro-fuzzy systems in
different way. An interface is developed to select an
appropriate type of the hybridization as shown in Fig. 5.
IV. CONCLUSION
The generic neuro-fuzzy framework presented here
facilitates automatic development of intelligent system in
friendly way. Using the framework many successful systems
can be developed. To name a few: course selection and career
advice system, student’s aptitude evaluation system, neuro-
fuzzy recruitment system, and portfolio management system.
It provides advantages such as reusability, modularity, and
flexibility. The system developed using this framework can be
saved for future use and served as good knowledge
management tool. As user need not have to write codes for the
components (just on demand attachment of component can be
done), user may concentrate on analysis and design of the
system in order to increase quality of the system. Obviously,
the framework reduces the total man power required to build
intelligent system, especially minimizes the role and need of
computer professionals. It will also save time and save manual
programming work. This generic library and framework are
developed using Microsoft’s Dot Net technology (Visual
Studio 2010) and this can be updated easily to future release of
versions. In future more components /codes can be added in to
the framework along with visual interface.
REFERENCES
[1] Zadeh, A.L. (1965). Fuzzy sets. Journal of Information and Control, 8(3),
338-353.
[2] Mamdani, E.H. (1993). Twenty years of fuzzy control: Experiences
gained and lessons learnt. In Proceedings of Second IEEE International
Conference on Fuzzy Systems, San Francisco, CA, 339-344.
[3] Takagi, T., & Sugeno, M. (1985). Fuzzy identification on systems and
its applications to modeling and control. IEEE Transactions on Systems,
Man, and Cybernetics, 15(1), 116-132.
[4] Lin, C,T., & Lee, C.S.G. (1991). Neural network based fuzzy logic
control and decision system. IEEE Transactions on Computer, 40(12),
1320-1336.
[5] Jang, R. (1992). Neuro-fuzzy modeling: Architectures, analyses and
applications. PhD Thesis, University of California, Berkeley.
[6] Bherenji, H.R., & Khedkar, P. (1992). Learning and tuning fuzzy logic
controllers through reinforcements. IEEE Transactions on Neural
Networks, 3(5), 724-740.
[7] Nauck, D., & Kruse, R. (1997). Neuro-fuzzy systems for function
approximation. In Proceedings of 4th International Workshop on Fuzzy-
Neuro Systems,Germany.
[8] Tano, S., Oyama, T., & Arnould, T. (1996). Deep combination of fuzzy
inference and neural network in fuzzy inference. Fuzzy Sets and
Systems, 82(2), 151-160.
[9] Sulzberger, S.M., Gurman, N.N., Vestli, S.J. (1993). Fun: Optimization
of fuzzy rule based systems using neural networks. In Proceedings of
IEEE Conference on Neural Networks, San Francisco, 312-316.
[10] Kasabov, N., & Qun, S. (1999). Dynamic evolving fuzzy neural
networks with 'm-out-of-n' activation nodes for on-line adaptive systems.
Technical Report TR99/04, Department of Information Science,
University of Otago.
[11] Feng, J.C., Teng, L.C. (1998). An online self constructing neural fuzzy
inference network and its applications. IEEE Transactions on Fuzzy
Systems, 6(1), 12-32.
[12] Cihan H. Dagli, Victor Pulla, Xavier Serrano, Modeling of a Neuro
Fuzzy System to Develop an Efficient Method to Get a Specific Color
Paint from the Color Model Cyan, Magenta and Yellow (CMY) under
Terms of Open Source, Procedia Computer Science, Volume 61, 2015,
Pages 486-491, ISSN 1877-0509 , Complex Adaptive Systems San Jose,
CA November 2-4, 2015, http://dx.doi.org/10.1016/j.procs.2015.09.196.
[13] Abraham, A. (2005). Adaptation of fuzzy inference system using neural
learning. StudFuzz 181, 53-83,
http://www.softcomputing.net/nf_chapter.pdf
[14] Jang, S.R., Sun, C.T., & Mizutani, E. (1997). Neuro fuzzy and soft
computing. Prentice Hall of India Ltd, India.
[15] Mendel, J.M. (2007). Type-2 fuzzy sets and systems: An Overview.
IEEE Computational Intelligence Magazine, 2(2), 20-29.
[16] Wu, H. & Mendal, J.M. (2002). Uncertainty bounds and their use in the
design of interval type-2 fuzzy logic system. IEEE Transactions on
Fuzzy Systems, 10(5), 622-639.
[17] Olivas, E.S., Guerrero, J.D., Valls, G.C., López, A.J., Maravilla, J.C. &
Chova, L.G. (2003). A low-complexity fuzzy activation function for
artificial neural networks. IEEE Transactions on Neural Networks, 14(6),
1576-1579.
[18] Su, C.L., Chen, C.J., & Yang, S.M. (2010). A self-organized neuro-
fuzzy system for stock market dynamics modeling and forecasting.
WSEAS Transactions on Information Science and Applications, 7(9),
1137-1149.
[19] John, R.I., & Coupland, S. (2007). Type-2 fuzzy logic: A historical view.
IEEE Computational Intelligence Magazine, 2(1), 57-62.
[20] Castillo, O., & Melin, P. (2008). Type 2 fuzzy logic: Theory and
applications. Springer.
[21] Malkawi, M., & Murad, O. (2013). Artificial neuro fuzzy logic system
for detecting human emotions. Human-Centric Computing and
Information Sciences, 3(3).
[22] Nie, D., Wang, X., Shi, L., & Lu, B. (2011). EEG-based emotion
recognition during watching movies. In Proceedings of International
IEEE EMBS Conference on Neural Engineering, Cancun, Mexico, 667-
670.
[23] Bouzaidaa, S., Saklya, A., & Sahlia, F.M. (2014). Extracting TSK-type
neuro-fuzzy model using the hunting search algorithm. International
Journal of General Systems, 43(1), 32-43.
[24] Azriyenni, & Mustafa, M.W. (2013). Performance neuro-fuzzy for
power system fault location. International Journal of Engineering and
Technology, 3(4), 497-501.
Priti Srinivas Sajja
Volume: 02, Issue: 01, January 2017
ISBN: 978-0-9957075-2-8
www.ijeacs.com
17
AUTHOR PROFILE
Priti Srinivas Sajja is a Professor at P. G.
Department of Computer Science,Sardar Patel
University, India since 1994. She specializes in
Artificial Intelligence, soft computing and
multiagent systems. She is co-author of
Intelligent Techniques for Data Science (2016);
Intelligent Technologies for Web Applications
(2012) and Knowledge-Based Systems (2009)
published at Switzerland and USA apart from
four books published in India. She is supervising
work of a few doctoral research scholars while
six candidates have completed their Ph.D. research under her guidance. She
was Principal Investigator of a major research project funded by UGC, India.
She has 171 publications in books, book chapters, journals, and in the
proceedings of national and international conferences out of which five
publications have won best research paper awards.
© 2017 by the author(s); licensee Empirical Research Press Ltd. United Kingdom. This is an open access article
distributed under the terms and conditions of the Creative Commons by Attribution (CC-BY) license.
(http://creativecommons.org/licenses/by/4.0/).
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
This paper is proposed for power systems fault location using Neuron Fuzzy (NF). The NF consists of approaching architecture Neural Network and Fuzzy Sets. The Neural Network is being responsible for detecting faults involving a limited number of components. The fuzzy sets represented diagram consisting of three node, that are; node1 for the system components, and node 2 for relays, node 3 for the circuit breakers. The NF uses primary and backup information to protective devices and to set generate training. The NF can be retained and estimated effectively. The software was conducted to show the effectiveness of the proposed NF is tested by using 13 bus test system.
Article
Full-text available
In this paper we present the modeling of a neuro-fuzzy system to be applied in the mix of paint from color model cyan, magenta and yellow (CMY). We aim to research an efficient method to get a specific hue color that is located within color wheel. The modeling is focused in to generate a wide gamma of colors to paint crafts like Panama Hats only with mentioned colors. Variables of the system are given by the frame and shape of selection of color. We use a fuzzy controller to generate the magnitude of the CMY percentages and a neural network to interpret the components to dosage the assigned color. Finally proceed data are used to develop an embedded controller under open source, the resulting color from the system is verified using different samples to determine the accuracy with selected color.
Article
Full-text available
This paper proposes a Takagi-Sugeno-Kang (TSK) type Neuro-Fuzzy model tuned by a novel metaheuristic optimization algorithm called Hunting Search (HuS). The HuS algorithm is derived based on a model of group hunting of animals such as lions, wolves, and dolphins when looking for a prey. In this study, the structure and parameters of the fuzzy model are encoded into a particle. Thus, the optimal structure and parameters are achieved simultaneously. The proposed method was demonstrated through modeling and control problems, and the results have been compared with other optimization techniques. The comparisons indicate that the proposed method represents a powerful search approach and an effective optimization technique as it can extract the accurate TSK fuzzy model with an appropriate number of rules.
Article
Full-text available
This paper presents an adaptive neuro/fuzzy system which can be trained to detect the current human emotions from a set of measured responses. Six models are built using different types of input/output membership functions and trained by different kinds of input arrays. The models are compared based on their ability to train with lowest error values. Many factors impact the error values such as input/output membership functions, the training data arrays, and the number of epochs required to train the model.
Conference Paper
Full-text available
This study aims at finding the relationship between EEG signals and human emotions. EEG signals are used to classify two kinds of emotions, positive and negative. First, we extracted features from original EEG data and used a linear dynamic system approach to smooth these features. An average test accuracy of 87.53% was obtained by using all of the features together with a support vector machine. Next, we reduced the dimension of features through correlation coefficients. The top 100 and top 50 subject-independent features were achieved, with average test accuracies of 89.22% and 84.94%, respectively. Finally, a manifold model was applied to find the trajectory of emotion changes.
Article
This paper presents an adaptive neuro/fuzzy system which can be trained to detect the current human emotions from a set of measured responses. Six models are built using different types of input/output membership functions and trained by different kinds of input arrays. The models are compared based on their ability to train with lowest error values. Many factors impact the error values such as input/output membership functions, the training data arrays, and the number of epochs required to train the model. ANFIS editor in MATLAB is used to build the models.
Article
A self-organized, five-layer neuro-fuzzy model is developed to model the dynamics of stock market by using technical indicators. The model effectiveness in prediction and forecasting is validated by a set of data containing four indicators: the stochastic oscillator (%K and %D), volume adjusted moving average (VAMA) and ease of movement (EMV) from TAIEX (Taiwan Stock Exchange Capitalization Weighted Stock Index). A modified moving average method is proposed to predict the input set for the neuro-fuzzy model in forecasting stock price. Simulation results show that the model is effective in prediction and accurate in forecasting. The input error from the prediction of the modified moving average method is attenuated significantly by the neuro-fuzzy model to yield better forecasting results.
Article
A mathematical tool to build a fuzzy model of a system where fuzzy implications and reasoning are used is presented. The premise of an implication is the description of fuzzy subspace of inputs and its consequence is a linear input-output relation. The method of identification of a system using its input-output data is then shown. Two applications of the method to industrial processes are also discussed: a water cleaning process and a converter in a steel-making process.
Chapter
The integration of neural networks and fuzzy inference systems could be formulated into three main categories: cooperative, concurrent and integrated neuro-fuzzy models. We present three different types of cooperative neurofuzzy models namely fuzzy associative memories, fuzzy rule extraction using self-organizing maps and systems capable of learning fuzzy set parameters. Different Mamdani and Takagi-Sugeno type integrated neuro-fuzzy systems are further introduced with a focus on some of the salient features and advantages of the different types of integrated neuro-fuzzy models that have been evolved during the last decade. Some discussions and conclusions are also provided towards the end of the chapter.