CH 4 DEEP LEARNING
1.Concept of Deep Learning
Concept of Deep Learning
6
Definition:
Deep Learning is a subset of Machine Learning (ML) in Artificial Intelligence
(AI) that uses artificial neural networks with multiple layers to model complex
patterns in data.
5
Key Points:
. Artificial Neural Networks (ANNs):
○ Deep learning is based on ANNs that mimic the human brain.
○ These networks consist of an input layer, hidden layers, and an
output layer.
. "Deep" means many hidden layers:
○ Traditional ML uses shallow networks (1-2 layers).
○ Deep learning uses multiple hidden layers, hence the term "deep".
. Learning from Data:
4
○ Deep learning systems automatically learn features from raw data
(e.g., images, audio, text).
○ It reduces the need for manual feature extraction.
. Popular Architectures:
3
○ Convolutional Neural Networks (CNNs) – used for image
processing.
○ Recurrent Neural Networks (RNNs) – used for sequential data like
2
time series and text.
○ Autoencoders – used for unsupervised learning and dimensionality
reduction.
○ Generative Adversarial Networks (GANs) – used for generating
1
new data.
. Training Process:
○ Uses backpropagation and gradient descent to adjust weights.
○ Requires large datasets and high computational power (GPUs/
TPUs).
. Applications:
○ Image and speech recognition
○ Natural Language Processing (NLP)
○ Self-driving cars
○ Healthcare diagnostics
○ Recommendation systems
7
. Advantages:
○ High accuracy with large datasets
○ Learns features automatically
○ Performs well in unstructured data (like images, audio)
. Disadvantages:
○ Requires a lot of data and computational resources
○ Difficult to interpret (black-box model)
○ Longer training time
2.Introduction to Neural Networks.
Introduction to Neural Networks
Definition:
A Neural Network is a computational model inspired by the structure and function
of the human brain. It is used in Machine Learning and Artificial Intelligence to
recognize patterns and solve complex problems.
Structure of a Neural Network:
. Neurons (Nodes):
○ The basic unit of a neural network.
○ Each neuron receives input, processes it, and passes the output to
the next layer.
. Layers:
○ Input Layer: Takes input data (features).
○ Hidden Layers: Perform calculations and extract patterns.
○ Output Layer: Produces the final prediction or classification.
. Weights and Biases:
○ Weights determine the strength of the connection between neurons.
○ Bias allows shifting the output curve.
. Activation Function:
○ Decides whether a neuron should be activated or not.
○ Common activation functions: Sigmoid, ReLU, Tanh, Softmax.
Working of Neural Network:
. Input data is passed through the input layer.
. Each neuron in hidden layers applies a mathematical function on inputs
(using weights, biases, and activation functions).
. Output is generated in the output layer.
. During training, the network learns by adjusting weights using
.
Backpropagation and Gradient Descent.
Types of Neural Networks:
● Feedforward Neural Network (basic type)
● Convolutional Neural Network (CNN) – used in image processing
● Recurrent Neural Network (RNN) – used in time-series and text
● Radial Basis Function Network (RBFN)
Applications:
● Image and speech recognition
● Natural language processing (chatbots, translation)
● Medical diagnosis
● Financial forecasting
Advantages:
● Can model complex non-linear relationships
● Learns from data automatically
● Works well with large datasets
Disadvantages:
● Requires a large amount of data and computation
● Acts like a “black box” (difficult to interpret)
● May overfit if not trained properly
3.Types of Deep Learning models.
Types of Deep Learning Models
Deep Learning models are built using different types of Artificial Neural
Networks (ANNs). Each model is designed for a specific type of task or data.
1. Feedforward Neural Network (FNN)
● Also called Multilayer Perceptron (MLP).
● Data flows in one direction—from input to output.
● No feedback or memory.
● Used for: Classification, regression.
2. Convolutional Neural Network (CNN)
● Specially designed for image and video processing.
● Uses convolutional layers to detect features like edges, textures, etc.
● Reduces the need for manual feature extraction.
Applications: Image classification, object detection, face recognition.
3. Recurrent Neural Network (RNN)
● Designed to handle sequential data (time-series, text, etc.).
● Has memory to store previous inputs using loops.
● Can suffer from vanishing gradient problem.
Applications: Language modeling, speech recognition, machine translation.
4. Long Short-Term Memory (LSTM)
● A special type of RNN that solves the vanishing gradient problem.
● Can remember data for long periods.
● More efficient than basic RNNs for long sequences.
Applications: Text generation, music generation, stock prediction.
5. Generative Adversarial Network (GAN)
● Consists of two models: Generator and Discriminator.
● Generator creates fake data; Discriminator tries to detect fake vs real.
● Used for generating new, realistic data.
Applications: Image generation, deepfake, style transfer.
6. Autoencoder
● An unsupervised model used for data compression and reconstruction.
● Learns to encode input data into a smaller representation and decode it
back.
Applications: Noise removal, anomaly detection, dimensionality reduction.
7. Transformer
● A powerful model for Natural Language Processing (NLP).
● Uses attention mechanism instead of recurrence.
● Basis for models like BERT, GPT, etc.
Applications: Text translation, summarization, chatbots.
Summary Table:
Model Type Main Use
FNN / MLP Classification, regression
CNN Image and video processing
RNN Sequential data (text, speech)
LSTM Long sequence memory tasks
GAN Data generation
Autoencoder Compression, noise removal
Transformer Natural Language Processing (NLP)
CNN Image and video processing
RNN Sequential data (text, speech)
LSTM Long sequence memory tasks
GAN Data generation
Autoencoder Compression, noise removal
Transformer Natural Language Processing (NLP)
4.Deep leaning applications.
Applications of Deep Learning
Deep Learning is widely used in various industries due to its ability to learn from
large amounts of data and solve complex problems. Below are some of the major
applications:
1. Image Recognition
● Identifies objects, people, scenes, and patterns in images.
● Used in:
○ Facial recognition (e.g., Face ID)
○ Medical imaging (e.g., tumor detection)
○ Security systems
2. Speech Recognition
● Converts spoken language into text.
● Used in:
○ Virtual assistants (e.g., Google Assistant, Siri, Alexa)
○ Voice-controlled systems
○ Call center automation
3. Natural Language Processing (NLP)
● Enables machines to understand and generate human language.
● Used in:
○ Chatbots
○ Machine translation (e.g., Google Translate)
○ Sentiment analysis (e.g., social media monitoring)
4. Autonomous Vehicles
● Deep learning helps self-driving cars make decisions using sensors and
cameras.
● Used in:
○ Lane detection
○ Obstacle detection
○ Traffic sign recognition
5. Healthcare
● Helps doctors diagnose diseases with better accuracy.
● Used in:
○ Analyzing X-rays and MRI scans
○ Predicting patient outcomes
○ Drug discovery
6. Finance
● Automates tasks and detects patterns in financial data.
● Used in:
○ Fraud detection
○ Credit scoring
○ Stock market prediction
7. Recommendation Systems
● Suggests content based on user behavior.
● Used in:
○ YouTube, Netflix, and Amazon recommendations
○ E-commerce platforms
8. Gaming and Simulation
● Creates realistic game characters and environments.
● Used in:
○ AI-powered NPCs (non-player characters)
○ Game development and testing
9. Agriculture
● Monitors crop health and improves yield.
● Used in:
○ Drone image analysis
○ Pest and disease detection
10. Robotics
● Helps robots understand their environment and act accordingly.
● Used in:
○ Human-following robots
○ Industrial automation
Summary:
Field Deep Learning Use
Healthcare Diagnosis, image analysis
Finance Fraud detection, forecasting
Automotive Self-driving cars
Retail Personalized recommendations
Agriculture Crop and soil monitoring
Security Face and voice recognition
