UNIT 1

SOFT COMPUTING
We use soft computing instead of (or alongside) hard computing in many cases because of the nature
of real-world problems.
Hard Computing (Traditional Computing)
Based on exact algorithms, deterministic systems.
Requires: Precise input and clear-cut rules.
Example: Arithmetic operations, database queries, exact sorting algorithms.
Soft Computing
Based on Approximate reasoning, uncertainty handling, learning from data.
Includes: Fuzzy logic, neural networks, genetic algorithms, probabilistic reasoning.
Goal: Provide good-enough, flexible, and robust solutions in complex or uncertain environments.
Why Use Soft Computing Instead of Hard Computing?
1. Incomplete or uncertain as soft computing can make decisions even with noisy, imprecise, or
incomplete data (e.g., fuzzy logic).
2. Real world complexity as many real-world problems are too complex for exact algorithms (e.g.,
weather prediction, speech recognition).
3. Human-like reasoning as Mimics the way humans make decisions—tolerant of imprecision, partial
truth, and uncertainty.
4. Adaptability as Learns and adapts over time (e.g., neural networks in AI systems).
5. Optimization problems as Effective for complex optimization problems where traditional methods
are inefficient (e.g., genetic algorithms).
Example
1. Problem recognition: Image recognition
Hard Computing Approach: Needs exact pattern matching (hard to scale).
Soft Computing Approach: Neural networks learn patterns from data
|
2. Problem recognition: Controlling an air conditioner
Hard Computing Approach: Uses fixed rules based on precise temperature
Soft Computing Approach: Fuzzy logic adjusts based on approximate comfort levels.

3. Problem recognition: Diagnosing diseases

Hard Computing Approach: Requires perfect symptom-to-diagnosis mapping
Soft Computing Approach: Probabilistic models infer diagnosis with uncertainty
Conclusion
We use soft computing when flexibility, learning, and approximation are more practical than precision.
It’s not a replacement for hard computing, but a complementary approach when real-world ambiguity
and complexity are involved.
Neural Network:
A Neural Network is a computational model inspired by the way biological neural networks in the
human brain process information. It is a subset of Artificial Intelligence (AI) and more specifically a
major part of Machine Learning (ML) and Deep Learning.
In simple terms, a neural network is made up of layers of nodes (neurons). Each neuron receives inputs,
processes them using weights and activation functions, and passes the output to the next layer. The
network learns from data by adjusting these weights through training, which enables it to make
predictions or decisions.
Why Neural Networks:
Neural Networks are used because they:
• Mimic the Human Brain: They try to replicate the way humans learn and make decisions.
• Handle Non-linear Relationships: Many real-world problems (like recognizing faces or
voices) have complex, non-linear patterns. Traditional algorithms may fail here, but neural
networks can model such complexities.
• Learn from Data: Neural networks can be trained on large volumes of data and can improve
their performance over time.
• Automation: They can automate tasks like classification, detection, and prediction without
explicitly programming them for each scenario.
Uses / Applications of Neural Networks:
Neural Networks are widely used in many domains. Some examples include:
a. Computer Vision
• Face recognition, object detection, image classification
• Used in social media, surveillance systems, and autonomous cars
b. Natural Language Processing (NLP)
• Language translation, chatbots, sentiment analysis, speech recognition
• Examples: ChatGPT, Google Translate, Alexa
c. Healthcare
• Diagnosis of diseases, analyzing X-rays, MRI scans, cancer detection
• Personalized treatment plans using patient data
d. Finance
• Stock market prediction, credit scoring, fraud detection
e. Autonomous Systems
• Self-driving cars use deep neural networks for vision, decision-making, and control
f. Gaming and Robotics
• Real-time decision making and behaviour modelling in games and robots
Advantages of Neural Networks:

Advantage Explanation

Neural networks learn from data and improve automatically without

1. Ability to Learn
human effort.

Suitable for processing images, audio, video, text, and other complex
2. Handle Complex Data
formats.

3. Non-linear Problem
Can model and solve problems that traditional algorithms cannot.
Solving

They can adapt to changing inputs and environments, making them very
4. Adaptability
flexible.

Once trained, they perform well even on unseen data, avoiding

5. Generalization
overfitting.

6. High Accuracy Often outperform traditional models when enough data is available.

7. Parallel Processing Neural networks can run on GPUs for high-speed parallel computation.

Neuron, Nerve Structure, and Synapse, which form the biological foundation for Artificial Neural
Networks (ANNs).

Neuron (Biological Neuron)

A neuron is the fundamental unit of the nervous system in humans and animals. It is a specialized cell
designed to transmit information throughout the body using electrical and chemical signals.

Structure of a Neuron:

A biological neuron typically consists of the following parts:

Part Description

Dendrites Short, branched projections that receive signals from other neurons.

Cell Body (Soma) Contains the nucleus and other organelles; processes incoming signals.

Axon A long fiber that transmits the signal from the neuron to other cells.

Axon Terminals Endpoints of the axon where signals are sent to other neurons.

Nerve Structure

A nerve is a bundle of axons from multiple neurons, enclosed in connective tissue. Nerves are part of
the peripheral nervous system and act as signal transmission highways.

Types of Nerves:

• Sensory Nerves – carry signals to the brain/spinal cord from the body.

• Motor Nerves – carry signals from the brain/spinal cord to muscles and glands.

• Mixed Nerves – contain both sensory and motor fibers.

Synapse

A synapse is the junction between two neurons, where information is transmitted from one neuron
to another.

Types of Synapses:

• Electrical Synapse – signal passes directly via ions through gap junctions.

• Chemical Synapse – signal is transmitted by neurotransmitters (most common in humans).

Importance in Artificial Neural Networks:

In Artificial Neural Networks (ANNs):

Biological Component ANN Equivalent

Neuron Artificial Neuron (Node)

Dendrites Inputs

Synapse Weights

Axon Output signal

Adjustable Weight (learned

Synaptic strength
during training)

Signal Transmission Activation Function + Output