Module 5

Hybrid System
Introduction
The main aim of the concept of hybridization is to overcome the weakness in one
technique. While applying it and bringing out the strength of the other technique to
find solution by combining them.
Every soft computing technique has particular computational parameters (e.g.,
ability to learn, decision making) which make them suited for a particular problem
and not for others.
It has to be noted that neural networks are good at recognizing patterns but they
are not good at explaining how they reach their decisions.
On the contrary, fuzzy logic is good at explaining the decisions but cannot
automatically acquire the rules used for making the decisions.
Also, the tuning of membership functions becomes an important issue in fuzzy
modelling.
Since this tuning can be viewed as an optimization problem, either neural network
(Hopfield neural network gives solution to optimization problem) or genetic
algorithms offer a possibility to solve this problem.
These limitations act as a central driving force for the creation of hybrid soft
computing systems where two or more techniques are combined in a suitable
manner that overcomes the limitations of individual techniques.
Neuro-Fuzzy Hybrid Systems
A neuro-fuzzy hybrid system (also called fuzzy neural hybrid), proposed by J. S. R. Jang, is a learning
mechanism that utilizes the training and learning algorithms from neural networks to find parameters of a
fuzzy system (i.e., fuzzy sets, fuzzy rules, fuzzy numbers, and so on).
It can also be defined as a fuzzy system that determines its parameters by processing data samples by
using a learning algorithm derived from or inspired by neural network theory.
Alternately, it is a hybrid intelligent system that fuses artificial neural networks and fuzzy logic by
combining the learning and connectionist structure of neural networks with human-like reasoning style of
fuzzy systems.
Neuro-fuzzy hybridization is widely termed as Fully Neural Network (FNN) or Neuro-Fuzzy System (NFS).
The human like reasoning style of fully systems is incorporated by NFS (the more popular term is used
henceforth) through the use of fuzzy sets and a linguistic model consisting of a set of lF-THEN fuzzy rules
Comparison of Fuzzy Systems with Neural Networks
From the existing literature, it can be noted that neural networks and fuzzy
systems have some things in common.
If there does not exist any mathematical model of a given problem, then neural
networks and fuzzy systems can be used for solving that problem (e.g:, pattern
recognition, regression, or density estimation).
This is the main reason for the growth of the intelligent computing techniques.
Besides having individual advantages, they do have certain disadvantages that
are overcome by combining both concepts.
When neural networks are concerned, if one problem is expressed by sufficient
number of observed examples then only it can be used.
These observations are used to train the black box.
Though no prior knowledge about the problem is needed extracting
comprehensible rules from a neural network's structure is very difficult.
A fuzzy system, on the other hand, does not need learning examples as prior
knowledge; rather linguistic rules are required.
Moreover, linguistic description of the input and output variables should be given.
If the knowledge is incomplete, wrong or contradictory, then the fuzzy system must
be runed.
Comparison of neural and fuzzy processing
Fuzzy processing
Neural processing
Mathematical model not necessary
Mathematical model not necessary

Learning can be done from scratch A prior knowledge is needed

There are several learning algorithms Learning is not possible

Black-box behaviour
Simple interpretation and implementation
Characteristics of Neuro-Fuzzy Hybrids
A fuzzy system-based NFS is trained by means of a data-driven learning method
derived from neural network theory.

This heuristic causes local changes in the fundamental fuzzy system.

At any stage of the learning process- before, during, or after- it can be

represented as a set of fuzzy rules.

For ensuring the semantic properties of the underlying fuzzy system, the learning
procedure is constrained.
An NFS approximates ann-dimensional unknown function, partly represented by
training examples.

Thus fuzzy rules can be interpreted as vague prototypes of the training data.

As shown in figure, an NFS is Inputs Outputs given by a three layer feed forward
neural network model.

It can also be observed that the first layer corresponds m the input variables, and
the second and third layers correspond to the fuzzy rules and output variables,
respectively.

The fuzzy sets are converted to (fuzzy) connection weights.

NFS can also be considered as a system of fuzzy rules wherein the system can be
initialized in the form of fuzzy rules based on the prior knowledge available. Some
researchers use five layers- the fuzzy sets being encoded in the units of the
second and the fourth layer, respectively. It is, however, also possible for these
models to be transformed into three-layer architecture.
Cooperative Neural Fuzzy Systems
In this type of system, the artificial neural network (ANN) and fuzzy system work independently
from each other.

The ANN attempts to learn the parameters from the fuzzy system.

Four different kinds of cooperative fuzzy neural networks are shown in Figure.

The FNN in Figure learns fuzzy set from the given gaining data.

This is done, usually, by fining membership functions with a neural network; the fuzzy sets then
being determined offline.

This is followed by their utilization m form the fuzzy system by fuzzy rules that are given, and
not learned.

The NFS in Figure determines, by a neural network, the fuzzy rules from the training data.

Here again, the neural networks learn offline before the fuzzy system is initialized. The rule
learning happens usually by clustering on self-organizing feature maps. There is also the
possibility of applying fuzzy clustering methods to obtain rules.
For the neuro-fuzzy model shown in Figure, the parameters of membership
function are learnt online, while the fuzzy system is applied.

This means that, initially, fuzzy rules and membership functions must be defined
beforehand.

Also, in order to improve and guide the learning step, the error has to be
measured.

The model shown in Figure determines the rule weights for all fuzzy rules by a
neural network.

A rule is determined by its rule weight-interpreted as the influence of a rule. They

are then multiplied with the rule output.
General Neuro·Fuzzy Hybrid Systems (General NFHS)
General Neuro-fuzzy hybrid systems (NFHS) resemble neural networks where a fuzzy
system is interpreted as a neural network of special kind.

The architecture of general NFHS gives it an advantage because there is no

communication between fuzzy system and neural network.

Figure illustrates an NFHS. In this figure the rule base of a fuzzy system is assumed to be a
neural network; the fuzzy sets are regarded as weights and the rules and the input and
output variables as Neurons.

The choice m include or discard Neurons can be made in the learning step.

Also, the fuzzy knowledge base is represented by the Neurons of the neural network; this
overcomes the major drawbacks of both underlying systems.
Membership functions expressing the linguistic terms of the inference rules should
be formulated for building a fuzzy controller.

However, in fuzzy systems, no formal approach exists to define these functions.

Any shape, such as Gaussian or triangular or bell shaped or trapezoidal, can be

considered as a membership function with an arbitrary set of parameters.

Thus for fuzzy systems, the optimization of these functions in terms of

generalizing the data is very important; this problem can be solved by using neural
networks.

Using learning rules, the neural network must optimize cite parameters by fixing a
distinct shape of the membership functions; for example, triangular. But regardless
of the shape of the membership functions, training data should also be available.
The Neuro fuzzy hybrid systems can also be modelled in an another method.

In cit's case, the training data is grouped into several clusters and each duster is
designed to represent a particular rule.

These rules are defined by the crisp data points and are not defined linguistically.

Hence a neural network, in this case, might be applied to train the defined dusters.

The resting can be carried out by presenting a random resting sample to the
trained neural network.

Each and every output unit will return a degree which extends to fit to the
antecedent of rule.
Adaptive Neuro-Fuzzy Inference System (ANFIS) in MATLAB

The basic idea behind this Neuro-adaptive learning technique is very simple.

This technique provides a method for the fuzzy modelling procedure and learn
information about a data set, in order to compute the membership function parameters
that best allow the associated fuzzy inference system to track the given input output data.

This learning method works similarly to that of neural networks.

ANFIS Toolbox in MATLAB environment performs the membership function parameter

adjustments. The function name used in activate this molbox in anfis. ANFIS toolbox can
be opened in MATLAB either at command line prompt or at Graphical User Interface.
Based on the given input-output dam set, ANFIS mol box builds a Fuzzy Inference System whose
membership functions are adjusted either using back Propagation network training algorithm or Adaline
network algorithm, which uses least mean square learning rule.

This makes the fuzzy system to learn from the data they model.

The Fuzzy Logic Toolbox function that accomplishes this membership function parameter adjustment is
called anfis. The acronym ANFIS derives its name from adaptive Neuro-fuzzy inference system.

The anfis function can be accessed either from the command line or through dte ANFIS Editor GUI.

Using a given input/output data set, the toolbox function anfis constructs a fuzzy inference system (FIS)
whose membership function parameters are adjusted using either a backPropagation algorithm alone or in
combination with a least squares type of method.

This enables fuzzy systems w learn from the data they are modeling.
FIS Structure and Parameter Adjustment A network-type structure similar to that of a neural network can be
used to interpret the input/output.

This structure maps inputs thtough input membership functions and associated parameters, and then through
output membership functions and associated parameters to outputs.

During the learning process, the parameters associated with the membership functions will change.

A gradient vector facilitates the computation (or adjustment) of these parameters, providing a measure of how
well the fuzzy inference system models the input/output data for a given set of parameters.

After obtaining the gradient vector, any of several optimization routines could be applied to adjust the
parameters for reducing some error measure (defined usually by the sum of the squared difference between
the accrual and desired outputs).

anfis makes use of either back-propagacion or a combination of adaline and back-propagation, for
membership function parameter estimation.
Genetic Fuzzy Hybrid and Fuzzy Genetic Hybrid Systems

Currently, several researches has been performed combining fuzzy logic and
genetic algorithms (GAs), and there is an increasing interest in the integration of
these two topics. The integration can be performed in the following two ways:

By the use of fuzzy logic based techniques for improving genetic algorithm
behavior and modelling GA components.

1. This is called fuzzy genetic algorithms (FGk).

2. 2. By the application of genetic algorithms in various optimization and search
problems involving fuzzy systems.
An FGA is considered as a genetic algorithm that uses techniques or tools based
on fuzzy logic to improve the GA behavior modelling.

It may also be defined as an ordering sequence of instructions in which some of

the instructions or algorithm components may be designed with tools based on
fuzzy logic.

For example, fuzzy operators and fuzzy connectives for designing genetic
operators with different properties, fuzzy logic control systems for controlling the
GA parameters according to some performance measures, stop criteria,
representation tasks, ere.
GAs are utilized for solving different fuzzy optimization problems.

For example, fuzzy flowshop scheduling problems, vehicle routing problems with
fuzzy duetime, fuzzy optimal reliability design problems, fuzzy mixed integer
programming applied m resource distribution, job-shop scheduling problem with
fuzzy processing time, interactive fuzzy satisfying method for multi-objective 0-1,
fuzzy optimization of distribution networks, etc.