Artificial intelligence (AI)
Artificial intelligence (AI) is a broad field of computer science that focuses on
creating intelligent machines that can perform tasks that typically require
human intelligence.
Ex. ChatGPT, Chatbots, Recommendation Engines- Netflix, Spotify, and Amazon.
Sub fields : Soft computing, Machine learning, Natural language processing,
Computer vision, Robotics, Expert systems.
Soft computing
• Deals with uncertain, imprecise, or incomplete data.
• It encompasses various techniques,
• fuzzy logic,
• neural networks,
• and evolutionary algorithms.
• Soft computing is used to solve problems that are difficult or
impossible to solve using traditional methods like
mathematical models and algorithms,
Fuzzy logic
Fuzzy logic is a mathematical framework for dealing with uncertainty by assigning
degrees of membership to objects or concepts.
Ex. Decision-making systems, control systems, and image processing.
Membership functions and fuzzy set
Membership function is a mathematical function used in fuzzy logic to represent
the degree of membership of a variable in a fuzzy set.
A fuzzy set is a set where elements have varying degrees of membership, rather
than just being a binary "yes or no" member of the set.
Neural network
• Type of machine learning model that is inspired by the
structure and function of the human brain.
• It consists of layers of interconnected nodes (neurons) that
process and transmit information to generate an output.
• layers, including an input layer, one or more hidden layers,
and an output layer.
ANFIS- Adaptive Neuro Fuzzy Inference
System
A type of hybrid intelligent system that combines the adaptability of
neural networks with the interpretability of fuzzy logic.
It combines back propagation and least-squares methods to adapt the
parameters of the fuzzy inference system.
ANFIS has been used in various applications, including control
systems, pattern recognition, and data analysis.
ANFIS Architecture
It is based on a five-layer feedforward neural network
architecture, where each neuron in the network performs a
specific function. The five layers are:
Input layer, Fuzzy layer, Normalization layer, Inference layer,
Output layer.
ANFIS Architecture
The Takagi-Sugeno (T-S) fuzzy model is a type of fuzzy
inference system that can be used for modeling complex
nonlinear systems.
The T-S fuzzy model consists of a set of fuzzy rules that
map the input variables to the output variable.
Each fuzzy rule in the T-S fuzzy model is expressed in the
form of a linear function of the input variables, where the
coefficients of the linear function are fuzzy sets.
ANFIS Architecture
Fuzzy rule base: The fuzzy rule base of the T-S fuzzy model
consists of a set of if-then rules that map the input variables
to the output variable. Each rule has the form:
If x is A and y is B and ... then z = f(x,y,...)
where A, B, ... are fuzzy sets, x, y, ... are the input variables,
z is the output variable, and f is a linear function.
ANFIS Architecture
Fuzzy membership functions: The fuzzy membership
functions of the T-S fuzzy model map the input variables to
fuzzy sets.
The most commonly used membership functions are
triangular, trapezoidal, and Gaussian.
Rule 1: If x is A1 and y is B1, then fuzzy set
f1=p1x+q1y+r1;
Rule 2: If x is A2 and y is B2, then fuzzyset f2=p2x+q2y+r2;
ANFIS Architecture
Figure 1: (a) A two inputs first order Takagi-Sugeno fuzzy model with two rules; (b) The equivalent ANFIS architecture
ANFIS Architecture
ANFIS Architecture
Representing Tsukamoto Fuzzy ModelS:
Tsukamoto Model by fuzzy rule formed as
Where x is input variable, y is input variable, A is
consequent variable, B is consequent variable, k is constant as
the consequent, z is output variable, z = f(x, y) is a crisp
function in the consequent. Usually, f(x, y) is a polynomial in the
input variables x and y Where x is input variable, y is input
variable, A is consequent variable, B
is consequent variable, z is output
variable, and C is variable’s value.
ANFIS Architecture
Representing Mamdani Fuzzy Model
For the Mamdani fuzzy inference system with max-min composition, a corresponding
ANFIS can be constructed if discrete approximations are used to replace the integrals in
the centroid defuzzification scheme introduced in here. However, the resulting ANFIS
is much more complicated than either TS ANFIS or Tsukamoto ANFIS. The extra
complexity in structure and computation of Mamdani ANFIS with max-min
composition does not necessarily imply better learning capability or approximation
power. If we adopt sum-product composition and centroid defuzzification for a
Mamdani fuzzy model, a corresponding ANFIS can be constructed easily based on
Theorem directly without using any approximation at all.
Hybrid Learning Algorithm:
ANFIS can be trained by a hybrid learning algorithm .
In the forward pass the algorithm uses least-squares method to identify the consequent
parameters on the layer 4.
• In the backward pass the errors are propagated backward and the premise parameters
are updated by gradient descent.
Hybrid Learning Rule:
Combines:
• the gradient rule (back propogation);
• the least squares estimate
Hybrid Learning Algorithm:
Back propogation
Cross – fertilize ANFIS
Interpolation of nodes through the basis function.
- Basis function -linear regression of nodes
Radial Basis Function Networks:
These are 3-layer networks that can approximate any continuous function through a
basis function expansion.
• The basis functions here (which are data dependent as earlier) exhibit some radial
symmetry.
• These networks have the so called perfect interpolation property.
Radial Basis Function Networks:
Eq.1
Radial Basis Function Networks:
A very popular model is the Gaussian RBF network
Here the output is written as
Eq.2
Eq.3
Radial Basis Function Networks:
Hence the problem is:
given N points, Xi, i = 1, ・ ・ ・ ,N in <m, find p „representative‟ points in
<m.
• This is the „clustering problem‟ This is a problem of forming the data into
p clusters.
• We can take the „cluster centers‟ to be the representative points.
Eq.4
Radial Basis Function Networks:
Eq.5
Coactive ANFIS - CANFIS
It yields advantages from non linear fuzzy rules.
-Multiple output anfis with nonlinear fuzzy rules
In CANFIS both NN and FIS play an active role in a effort to reach a specific goal
Canfis has extended the notion of single-output system of ANFIS to produce multiple
outputs.
One way to accomplish is to place as many ANFIS models side by side as the number
of required outputs.
CANFIS -Architecture
The architecture of coactive neuro-fuzzy modeling (CNFM) typically consists of three
main components:
The neuro-fuzzy system,
The coactive modeling framework,
and the interaction loop between the human operator and the model.
Neuro-Fuzzy System:
Fuzzification Module, Inference Engine, Defuzzification Module
Coactive Modeling Framework:
Human Operator Interface, Learning and Adaptation Mechanism, Model Integration
and Control, Interaction Loop:Data Collection, Model Processing, Feedback and
Guidance
CANFIS
Neuron Functions for Adaptive networks
-Neurons play a crucial role in processing and transmitting information.
Activation function and Transfer function
The activation function determines the output of a neuron based on its input.
-It introduces non-linearity into the network and allows the neuron to model complex
relationships between inputs and outputs.
Some commonly used activation functions include:
Sigmoid Function- maps the input to a value between 0 and 1
Hyperbolic Tangent (tanh) Function - maps the input to a value between -1 and 1
Rectified Linear Unit (ReLU) - returns the input as the output if it is positive;
otherwise, it outputs zero.
Neuron Functions for Adaptive networks
Transfer Function- Determines how the neuron's input is transformed into an
output.
Linear Transfer Function - Linear regression and perceptron models.,
Radial Basis Function (RBF)- Radial basis kernel to transform the input into a higher-
dimensional space,
Gaussian Function-applies a Gaussian distribution to the input, transforming it based on
the mean and variance,
Learning Rule- determines how the neuron's weights or parameters are updated during
the learning process,
Hebbian Learning Rule- modifies the weights of connected neurons based on the
correlation between their activities,
Delta Rule (Widrow-Hoff Rule) adjusts the weights of a neuron based on the difference
between the desired output and the actual output,
Backpropagation- It computes the gradient of the network's error with respect to the
weights and updates them using gradient descent.
The specific choice of functions depends on the nature of the problem being solved and
the architecture of the network. Different functions can have different impacts on the
network's learning capacity, convergence speed, and generalization ability.
Neuro Fuzzy Spectrum
Spectrum refers- continuous sequence of function based on neural network and fuzzy
system.
It is a conceptual framework that represents the relationship between fuzzy logic and
neural networks, highlighting the different levels of fuzziness and neural network
characteristics.
How hybrid neuro-fuzzy systems can be positioned within this spectrum.
The Neuro-Fuzzy Spectrum categorizes these hybrid systems into three main levels:
1. Fuzzy Systems with Neural Network Learning,
2. Neural Networks with Fuzzy Logic Interpretation,
3. Fully Integrated Neuro-Fuzzy Systems.
The Neuro-Fuzzy Spectrum highlights the trade-offs and synergies between fuzzy logic
and neural networks, offering a framework to design and position hybrid neuro-fuzzy
systems based on the problem requirements, interpretability needs, data availability, and
desired system characteristics.
THANK YOU
