SNS COLLEGE OF ENGINEERING
Kurumbapalayam(Po), Coimbatore – 641 107
Accredited by NAAC-UGC with ‘A’ Grade
Approved by AICTE, Recognized by UGC & Affiliated to Anna University, Chennai
Department of AI &DS
Course Name – 19AD602 DEEP LEARNING
III Year / VI Semester
Unit 1-INTRODUCTION
Topic: INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE
INTRODUCTION TO DEEP LEARNING
Case study:
Identifying Handwritten Digits with Deep Learning
Context
The MNIST dataset consists of handwritten digits (0–9), widely used for training image processing systems. The challenge is to classify images of
these digits using a deep learning approach.
Objective
Build a neural network model to classify handwritten digits.
Approach
1. Data Collection: Use the MNIST dataset, which contains 60,000 training and 10,000 testing grayscale images. Each image is 28x28 pixels.
2. Model Architecture: A simple neural network with one input layer, one hidden layer with 128 neurons (ReLU activation), and an output
layer with 10 neurons (softmax activation for classification).
3. Training: Train the model using cross-entropy loss and the Adam optimizer.
4. Evaluation: Check the accuracy on the test set and visualize misclassified images.
Outcome
The model achieves approximately 97% accuracy on the test set, demonstrating the effectiveness of neural networks in recognizing handwritten
digits.
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 2/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 3/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 4/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 5/23
INTRODUCTION TO DEEP LEARNING
APPLICATIONS OF DEEP LEARNING:
1. Fraud detection
2. Customer service
3. Financial services
4. Natural language processing
5. Facial recognition
6. Self-driving vehicles
7. Predictive analytics
8. Recommender systems
9. Industrial…etc..,
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 5/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 7/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 8/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 9/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 10/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 11/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 12/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 13/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 14/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 15/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 16/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 17/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 18/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 19/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 20/23
INTRODUCTION TO DEEP LEARNING
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 21/23
INTRODUCTION TO DEEP LEARNING
Activity: Build Your Own Neural Network
Objective: Learn how a neural network operates by building one yourself.
Setup:
● Divide participants into small groups.
● Provide each group with a set of cards representing nodes and weights. Use colored markers to indicate inputs, hidden
layers, and outputs.
Steps:
1. Define the Input: Assign groups a task, e.g., classify whether a given "animal image" is a cat or dog (use analogies for
simplicity).
2. Build Layers: Ask groups to connect "input cards" (features) to "hidden nodes" and finally to "output cards" (class
labels).
3. Forward Pass: Groups calculate the outputs at each layer, mimicking matrix operations.
4. Activation Function: Introduce rules like "output > 0 means activate the node" (ReLU) and apply these rules.
5. Output Result: Summarize outputs for classification.
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 22/23
INTRODUCTION TO DEEP LEARNING
THANK YOU
GULSHAN BANU.A/ AP/AI AND DS / INTRODUCTION TO DEEP LEARNING/SNSCE 23/23