Introduction to Deep Learning

 

Introduction to Deep Learning

1. Introduction 


Question 1

How does your phone recognize your face?

Question 2

How does ChatGPT answer your questions?

Question 3

How does YouTube know which video you will watch next?

Question 4

How does Google Lens identify objects?

All these applications are powered by Deep Learning, one of the most influential technologies in Artificial Intelligence today.

2. What is Deep Learning? 

Definition

Deep Learning is a subset of Machine Learning that uses Artificial Neural Networks (ANNs) with multiple hidden layers to automatically learn complex patterns from large amounts of data.

Break the definition

Deep

  • Refers to multiple hidden layers in the neural network.

Learning

  • The model improves its performance by learning patterns from data.

Artificial Neural Network

  • A computational model inspired by the human brain.


Board Diagram

Artificial Intelligence (AI)

        ↓

Machine Learning (ML)

        ↓

Deep Learning (DL)

        ↓

Artificial Neural Networks (ANN)

Characteristics of Deep Learning

  • Learns automatically from data

  • Uses multiple hidden layers

  • Handles large datasets

  • Learns complex patterns

  • Requires high computational power

  • Suitable for unstructured data


Types of Data Used

Structured Data

  • Student marks

  • Salary

  • Bank transactions

Semi-Structured Data

  • XML

  • JSON

Unstructured Data

  • Images

  • Videos

  • Audio

  • Text

  • Speech


Machine Learning performs well on structured data.

Deep Learning excels with unstructured data.


3. Historical Trends in Deep Learning 

1943 – Artificial Neuron

Scientists:

Warren McCulloch and Walter Pitts

Contribution:

  • Proposed the first mathematical model of an artificial neuron.

  • This became the foundation of Artificial Neural Networks.

Real-world analogy:

Like inventing the first bicycle before modern cars.


1950 – Turing Test

Scientist:

Alan Turing

Contribution:

Asked a simple question:

"Can a machine think?"

Introduced the Turing Test to evaluate machine intelligence.

Example:

Today's ChatGPT is often discussed in relation to this idea.


1956 – Birth of Artificial Intelligence

At the Dartmouth Conference:

John McCarthy introduced the term Artificial Intelligence.

This is considered the birth of AI as a research field.


1958 – Perceptron

Scientist:

Frank Rosenblatt

Contribution:

Developed the Perceptron, the first Artificial Neural Network capable of learning simple patterns.

Limitation:

Could not solve non-linear problems such as XOR.


1969 – AI Winter Begins

Scientists:

Marvin Minsky and Seymour Papert

Contribution:

Published research showing the limitations of single-layer perceptrons.

Result:

Funding for AI research declined.

This period is known as the AI Winter.


1986 – Backpropagation

Scientists:

David Rumelhart, Geoffrey Hinton, and Ronald Williams

Contribution:

Introduced the Backpropagation algorithm for training multi-layer neural networks efficiently.

Importance:

This made training deeper networks practical.


1997 – IBM Deep Blue

IBM's Deep Blue defeated world chess champion Garry Kasparov.

This demonstrated that AI could solve highly complex problems.


2006 – Deep Learning Revival

Scientist:

Geoffrey Hinton

Contribution:

Showed effective methods for training deep neural networks.

He is often called the Father of Deep Learning.


2012 – AlexNet Revolution

Scientists:

Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton

Contribution:

AlexNet won the ImageNet competition by a large margin using deep learning.

Impact:

Deep Learning became the dominant approach in computer vision.


2016 – AlphaGo

Google DeepMind's AlphaGo defeated Go champion Lee Sedol.

This showed Deep Learning could tackle problems previously considered too complex for machines.


2017 – Transformer Architecture

Scientists:

Ashish Vaswani and colleagues

Paper:

Attention Is All You Need

Contribution:

Introduced the Transformer architecture.

This transformed Natural Language Processing.


2022 – ChatGPT

OpenAI introduced ChatGPT.

Applications:

  • Education

  • Coding

  • Research

  • Content Writing

  • Customer Support


Timeline on the Board

1943 → Artificial Neuron

1950 → Turing Test

1956 → Birth of AI

1958 → Perceptron

1969 → AI Winter

1986 → Backpropagation

1997 → IBM Deep Blue

2006 → Deep Learning Revival

2012 → AlexNet

2016 → AlphaGo

2017 → Transformers

2022 → ChatGPT

4. Why is Deep Learning Growing? 


If Deep Learning was invented many years ago, why did it become popular only recently?

Reason 1: Big Data

Earlier:

Thousands of records

Today:

Billions of records from:

  • Facebook

  • Instagram

  • YouTube

  • Amazon

  • Hospitals

  • Banks

  • IoT devices


Reason 2: High-Performance Hardware

Earlier:

Central Processing Units (CPUs)

Today:

Graphics Processing Units (GPUs)

Examples:

  • NVIDIA

  • AMD

GPUs process thousands of calculations simultaneously, making Deep Learning training much faster.


Reason 3: Cloud Computing

Cloud platforms provide scalable computing power.

Examples:

  • Google Cloud

  • Amazon Web Services (AWS)

  • Microsoft Azure


Reason 4: Better Algorithms

Examples:

  • Backpropagation

  • Rectified Linear Unit (ReLU)

  • Adam Optimizer

  • Batch Normalization

These algorithms improved training speed and accuracy.


Reason 5: Open-Source Frameworks

Examples:

  • TensorFlow

  • PyTorch

  • Keras

These frameworks made Deep Learning accessible to researchers and developers.


Reason 6: Better Storage

Organizations can now store enormous amounts of data at low cost.

Examples:

  • Cloud Storage

  • Data Lakes


Reason 7: Demand for AI Applications

Industries require AI for:

  • Healthcare

  • Banking

  • Education

  • Agriculture

  • Cybersecurity

  • Manufacturing

  • Transportation

  • Retail

  • Entertainment


Real-World Applications

Healthcare

  • Cancer detection from MRI scans

  • Disease diagnosis

  • Medical image analysis

Banking

  • Fraud detection

  • Credit risk analysis

Education

  • Intelligent tutoring systems

  • Automatic grading

  • Personalized learning

Retail

  • Product recommendation

  • Demand forecasting

Transportation

  • Self-driving cars

  • Traffic prediction

Cybersecurity

  • Malware detection

  • Intrusion detection

Agriculture

  • Crop disease detection

  • Smart irrigation

Manufacturing

  • Predictive maintenance

  • Quality inspection

Media & Entertainment

  • Movie recommendations

  • Music recommendations

  • AI-generated content

Natural Language Processing

  • ChatGPT

  • Translation systems

  • Virtual assistants

Computer Vision

  • Face recognition

  • Object detection

  • Autonomous drones


5. Introduction to Artificial Neural Networks (ANN) 

Transition:

"We now know why Deep Learning is powerful. But what is the engine behind Deep Learning?"

Answer:

Artificial Neural Network (ANN).

Definition

An Artificial Neural Network is a computational model inspired by the structure and functioning of the human brain. It consists of interconnected artificial neurons that process information, learn patterns from data, and make predictions or decisions.


Human Brain vs Artificial Neural Network

Human Brain        Artificial Neural Network
Biological Neuron      Artificial Neuron
Synapse                      Weight
Dendrites    Inputs
Cell Body    Processing Unit
Axon    Output

Explain that the brain has approximately 86 billion biological neurons, while an ANN uses mathematical neurons connected by weighted links to simulate learning.




Basic Structure of an ANN




  • Input Layer: Receives raw data (image pixels, text, numbers, etc.).

  • Hidden Layer(s): Learn patterns and relationships from the data.

  • Output Layer: Produces the final prediction or classification.


Real-Life Analogy

Suppose you are identifying a student.

  • Input: Face image

  • Hidden Layer 1: Detects edges and shapes

  • Hidden Layer 2: Detects eyes, nose, and mouth

  • Hidden Layer 3: Recognizes the complete face

  • Output: Student identified

This illustrates how an ANN learns progressively from simple features to complex representations.


Homework

Identify one real-world application of Deep Learning in any domain (healthcare, banking, agriculture, education, cybersecurity, retail, etc.) and describe:

  1. The problem being solved.

  2. The input data used.

  3. Why Deep Learning is preferred over traditional Machine Learning.

  4. The expected output.


टिप्पणी पोस्ट करा

0 टिप्पण्या