What are the main topics covered in the deep learning course?

The deep learning course covers several key topics, including the architecture of deep neural networks, training algorithms, and various applications. Students will learn about convolutional networks, recurrent neural networks, and autoencoders. Additionally, the course introduces generative adversarial networks and their applications in fields like computer vision and speech recognition. Each module is designed to build a comprehensive understanding of deep learning principles and practices.

What practical experiments are included in the deep learning syllabus?

The syllabus includes hands-on experiments such as building and training a feedforward neural network for classification tasks. Students will also create convolutional neural networks for image classification using datasets like CIFAR-10 or MNIST. Other experiments involve implementing recurrent neural networks for sequence prediction and developing autoencoders for image compression. These practical applications enhance the learning experience by providing real-world context.

How does the course address the applications of deep learning?

The course emphasizes the applications of deep learning across various domains, including computer vision, speech recognition, and natural language processing. Students will explore how deep learning techniques can be applied to solve complex problems in these fields. The syllabus introduces large-scale deep learning concepts and discusses the impact of deep learning on industry and innovation. This focus prepares students for careers in data science and artificial intelligence.

What is the significance of recurrent neural networks in deep learning?

Recurrent neural networks (RNNs) are crucial for processing sequential data, making them essential in deep learning applications like time series prediction and natural language processing. The course covers the architecture of RNNs, including Long Short-Term Memory (LSTM) networks, which help mitigate issues like vanishing gradients. Understanding RNNs allows students to tackle problems involving sequences, such as language modeling and speech recognition, effectively.

What foundational concepts are introduced in the deep learning course?

Foundational concepts in the deep learning course include understanding the importance of deep learning and its various architectures. Students will learn about activation functions, weight initialization, and regularization techniques that are vital for training deep neural networks. The course also discusses the building blocks of deep learning, enabling students to grasp the underlying principles that govern neural network performance.

Fundamentals of Deep Learning Course Syllabus

Key Points

Covers foundational concepts and advanced topics in deep learning.
Includes modules on convolutional networks and recurrent neural networks.
Features hands-on experiments with feedforward neural networks and GANs.
Explores applications of deep learning in computer vision and NLP.

24CSEN2331 FUNDAMENTALS OF DEEP LEARNING

L T P S J C

2 0 2 0 0 3

Pre-requisite 00000000

Co-requisite None

Preferable Exposure None

Course Description:

This course offers an in-depth exploration of deep learning, a subfield of machine learning focused on neural networks with

multiple layers. Students will delve into the architecture, training algorithms, and applications of deep neural networks. The

course covers both foundational concepts and advanced topics in deep learning, enabling students to build and apply deep

learning models to various real-world tasks.

Course Educational Objectives:

• To introduce the principles and significance of deep learning.

• To familiarize the student with the architecture and components of deep neural networks.

• To teach the applications of recurrent neural networks

• To edify autoencoders and deep generative models

• To create awareness of different applications of deep learning.

MODULE 1 INTRODUCTION TO DEEP LEARNING 9 Hrs

Importance and applications of deep learning, Overview of deep neural network architectures, Building blocks of deep neural

networks, Activation functions: sigmoid, ReLU, etc. Weight initialization and regularization techniques.

MODULE 2 CONVOLUTION NETWORKS 9 Hrs

Architectures, Convolution operations, Pooling layer, Variants of the basic Convolution Function, Efficient Convolution

algorithms.

MODULE 3 RECURRENT NEURAL NETWORKS (RNNS) AND SEQUENCES 9 Hrs

Understanding sequential data and RNNs, Long Short-Term Memory (LSTM) networks, Applications of RNNs.

MODULE 4 AUTO ENCODERS 9 Hrs

Auto encoders: Under complete auto encoders, regularized encoders, stochastic encoders and decoders.

MODULE 5 APPLICATIONS OF DEEP LEARNING 9 Hrs

Large scale Deep learning, Computer vision, speech recognition, NLP, and other applications. Introduction to Generative

Adversarial Networks (GANs) and their applications

List of Experiments

S.no Topic Type

Building and Training a Feedforward Neural Network Implement a basic feedforward neural

network for a classification task using either MATLAB or Python with a deep learning library

like TensorFlow or PyTorch.

Experiment

Convolutional Neural Network (CNN) for Image Classification Create and train a CNN for

image classification using a popular dataset (e.g., CIFAR-10 or MNIST) with appropriate data

augmentation techniques.

Experiment

Recurrent Neural Network (RNN) for Sequence Prediction Implement an RNN for sequence

prediction using LSTM cells. Train the model on a relevant dataset.

Experiment

Autoencoder for Image Compression Build and train an autoencoder to compress and

reconstruct images. Evaluate the quality of reconstruction and the compression ratio.

Experiment

Generative Adversarial Network (GAN) for Image Generation Develop a GAN to generate

realistic images. Train the generator and discriminator on a dataset such as CelebA or MNIST.

Experiment

Text Classification with Word Embeddings Perform text classification using word embeddings

and an RNN. Use a dataset with labeled text documents.

Experiment

LSTM for Time Series Prediction Apply LSTM networks for time series prediction, such as

predicting stock prices or temperature trends, using historical data.

Experiment

Speech Recognition using Convolutional Neural Networks Develop a speech recognition

system using a CNN. Train the model on a dataset of spoken words or phrases.

Experiment

Textbook(s):

1. Ian Goodfellow, Yoshua Bengio, Aaron Courville, Deep Learning, MIT Press, 2016 ,978-0262035613

2. Michael Nielsen, Neural Networks and Deep Learning, Determination Press, 2015

Reference(s):

1. Amlan Chakrabarti Amit Kumar Das, Saptarsi Goswami, Pabitra Mitra, Deep Learning, 1st Edition, Pearson,

2. Sandro Skansi, Introduction to Deep Learning, Springer,

Course Outcomes:

1. Understand the basic principles of deep learning (L2)

2. Explain the architecture and components of deep neural networks (L2).

3. Illustrate Recurrent Neural Networks (RNNs) and Sequences (L2)

4. Explain different autoencoders and deep generative models (L2)

5. Develop and evaluate deep learning models for various applications (L6)

Course Articulation Matrix:

POs PSOs

CO 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 2 3 4

1 2 1 2 1 1 1 1 2 3

2 2 1 2 1 1 1 3 2 1

3 2 1 3 1 1 1 3 2 2

4 2 1 3 1 1 1 3 2 2

5 2 1 2 1 1 1 3 2 2

3 – High, 2 – Medium & 1 – Low Correlation

APPROVED IN MEETINGS HELD ON:

BOS : 03-02-2024 Academic Council Number: 27 Academic Council : 06-07-2023

SDG No(s). & Statement(s) :

4 & Quality Education : Ensure inclusive and equitable quality education and promote lifelong learning opportunities for

all.Ensure inclusive and equitable quality education and promote lifelong learning opportunities for all.

9 & Industry, Innovation and Infrastructure : Build resilient infrastructure, promote inclusive and sustainable industrialization and

foster innovation.

SDG Justification(s):

SDG 4: The modules and topics mentioned in this course are designed to ensure all-inclusive and thorough education with

equity to all persons and always promote learning opportunities. SDG 9: The modules and topics mentioned in this course are

designed to ensure the engineers build resilient infrastructure which promote inclusive and sustainable industrialization and

foster innovation.

Overview

Fundamentals of Deep Learning Course Syllabus

Fundamentals of Deep Learning explores neural networks and their applications in various fields. The course covers deep learning architectures, training algorithms, and real-world applications, including computer vision and natural language processing. Students will learn about recurrent neural networks, autoencoders, and generative models. This syllabus outlines key modules and experiments designed for students interested in machine learning and artificial intelligence. Ideal for those pursuing careers in data science and AI development. Key Points Covers foundational concepts and advanced topics in deep learning. Includes modules on convolutional networks and recurrent neural networks. Features hands-on experiments with feedforward neural networks and GANs. Explores applications …