Deep Learning

Introduction to Deep Learning

• Deep Learning is a subfield of Machine
Learning (ML) that focuses on using neural
networks with multiple layers to model
complex patterns in data.
• It is inspired by the structure and functioning
of the human brain, using artificial neurons to
process information and learn from large
datasets.
Why Deep learning is becoming popular?

• Data Growth- Volume of data has grown u in

terms of business, personal dat and so on.
• Advancement in hardware – GPU, TPU
• Python and Opensource Ecosystem – Python a
easy programming language.
• Cloud - run a server on cloud. No need of
hardware in the system.
Why Deep Learning?
• Deep Learning has gained popularity due to its
ability to:
✅ Learn features automatically from raw data
(eliminating the need for manual feature extraction).
✅ Scale well with large datasets and computational
power.
✅ Achieve state-of-the-art performance in tasks like
image recognition, natural language processing
(NLP), speech recognition, and reinforcement
learning.
Difference Between AI, ML, and DL
• AI: Broad field for simulating human
intelligence.
• ML: Algorithms that learn from data (subset of
AI).
• DL: Uses multi-layered neural networks
(subset of ML).
Deep Learning Fundamentals

• Deep Learning is a subset of Machine

Learning (ML) that uses Artificial Neural
Networks (ANNs) to automatically learn
patterns from large datasets.
• It has revolutionized fields like computer
vision, natural language processing, and
speech recognition by achieving human-level
performance in many tasks.
Artificial Neural Networks (ANNs)
• Structure of a Neural Network
A typical Feedforward Neural Network (FNN)
consists of:
• - Input Layer: Receives raw data.
• - Hidden Layers: Learn complex patterns.
• - Output Layer: Provides predictions.
Each neuron applies a weighted sum followed by
an activation function to decide whether to pass
the signal forward.
Activation Functions
Activation functions introduce non-linearity,
allowing neural networks to learn complex
relationships. Common activation functions include:
• - ReLU: Most common, replaces negatives with 0.
• - Sigmoid: Outputs between 0 and 1, used for
classification.
• - Tanh: Outputs between -1 and 1.
• - Softmax: Used for multi-class classification.
Forward & Backpropagation
Neural networks learn by adjusting weights using two
key steps:
1. Forward Propagation
1. Input data moves through the network layer by layer.
2. Final output is compared with the actual label using a loss
function.
2. Backpropagation
1. Computes the error gradient (how much each weight
contributes to the error).
2. Uses Gradient Descent to update the weights and minimize
error.
Loss Functions & Optimization
To improve predictions, Deep Learning models minimize a loss
function using optimization algorithms.
Common Loss Functions
📉 Mean Squared Error (MSE) – Used for regression problems.
📉 Cross-Entropy Loss – Used for classification problems.
Optimization Algorithms
• Gradient Descent (GD) – Iteratively adjusts weights based on
the loss gradient.
Stochastic Gradient Descent (SGD) – Updates weights using
random samples; faster but noisier.
Adam (Adaptive Moment Estimation) – Combines advantages
of GD and SGD; widely used.
Deep Learning Architectures
Deep Learning is powerful due to its varied architectures
designed for different types of data.
Key Architectures
• 📷 Convolutional Neural Networks (CNNs) – Specialized
for image processing.
📝 Recurrent Neural Networks (RNNs) & LSTMs – Handle
sequential data like text and speech.
🧠 Transformers (e.g., BERT, GPT) – Power modern NLP
models.
Deep Reinforcement Learning (DRL) – Used in AI agents
(e.g., AlphaGo, self-driving cars).
Types of neural networks

1. Feedforward neural networks (FNNs) are the simplest type

of ANN, where data flows in one direction from input to output.
It is used for basic tasks like classification.

2. Convolutional Neural Networks (CNNs) are specialized for

processing grid-like data, such as images. CNNs use
convolutional layers to detect spatial hierarchies, making them
ideal for computer vision tasks.

3. Recurrent Neural Networks (RNNs) are able to process

sequential data, such as time series and natural language.
RNNs have loops to retain information over time, enabling
applications like language modeling and speech recognition.
Variants like LSTMs and GRUs address vanishing gradient issues.
Types of neural networks

4. Generative Adversarial Networks (GANs) consist of two networks—a

generator and a discriminator—that compete to create realistic data.
GANs are widely used for image generation, style transfer, and data
augmentation.

5. Autoencoders are unsupervised networks that learn efficient data

encodings. They compress input data into a latent representation and
reconstruct it, useful for dimensionality reduction and anomaly detection.

6. Transformer Networks has revolutionized NLP with self-attention

mechanisms. Transformers excel at tasks like translation, text generation,
and sentiment analysis, powering models like GPT and BERT.
Challenges in Deep Learning

1. Data availability: It requires large amounts of data to learn

from. For using deep learning it’s a big concern to gather as much
data for training.
2. Computational Resources: For training the deep learning
model, it is computationally expensive because it requires
specialized hardware like GPUs and TPUs.
3. Time-consuming: While working on sequential data depending
on the computational resource it can take very large even in days
or months.
4. Interpretability: Deep learning models are complex, it works like
a black box. it is very difficult to interpret the result.
5. Overfitting: when the model is trained again and again, it
becomes too specialized for the training data, leading to
overfitting and poor performance on new data.
Advantages of Deep Learning

1. High accuracy: Deep Learning algorithms can achieve state-of-

the-art performance in various tasks, such as image recognition
and natural language processing.
2. Automated feature engineering: Deep Learning algorithms
can automatically discover and learn relevant features from data
without the need for manual feature engineering.
3. Scalability: Deep Learning models can scale to handle large and
complex datasets, and can learn from massive amounts of data.
4. Flexibility: Deep Learning models can be applied to a wide range
of tasks and can handle various types of data, such as images,
text, and speech.
5. Continual improvement: Deep Learning models can continually
improve their performance as more data becomes available.
Applications of Deep Learning
Deep Learning is used across industries:
• 🎭 Computer Vision – Face recognition, object
detection, medical imaging.
Natural Language Processing (NLP) – Chatbots,
sentiment analysis, translation.
🔊 Speech Recognition – Virtual assistants like Alexa,
Siri.
🚗 Autonomous Vehicles – Self-driving cars use CNNs &
Reinforcement Learning.
📊 Finance & Healthcare – Fraud detection, stock
predictions, medical diagnostics.
Tools & Frameworks
• TensorFlow – Developed by Google; used for
production-scale models.
• PyTorch – Developed by Facebook; popular for
research and experimentation.
• Keras – High level API built on TensorFlow for
easier model building.
How to Get Started?
• 1. Learn Python (NumPy, Pandas, Matplotlib).
• 2. Build simple models using
TensorFlow/PyTorch.
• 3. Work on small projects (digit classification,
sentiment analysis).
• 4. Explore pre-trained models (VGG, ResNet,
BERT).