MEPCO R2023 ME COMPUTER SCIENCE AND ENGINEERING

23MC202 DEEP LEARNING TECHNIQUES L T P C

Offered by CSE (Core Theory) 3 0 0 3
Course Objective:
To impart knowledge on
To learn the basic concepts of Machine learning.
To understand convolutional networks.
To understand sequence modelling.
To study Auto encoders and Generative adversarial networks.
To work with real world problems.
Course Outcome:
At the end of the course, the students will be able to
Implement feed forward neural networks.
Apply convolutional networks for problem solving.
Solve problems using sequence modelling.
Work with Auto encoders and Generative adversarial networks.
Apply Deep neural networks in various real time applications.
Unit I MACHINE LEARNING BASICS 9
Scalars, Vectors, Matrices and Tensors Numerical computation: overflow and underflow Machine learning tasks
Computational model of a neuron - Feed forward neural networks - Overfitting and Underfitting Gradient
based learning Early stopping Challenges in Neural network optimization - Algorithms with adaptive learning
rates: AdaGrad, RMSProp, Adam, Choosing the Right Optimization algorithm.
Unit II CONVOLUTIONAL NETWORKS 9
Convolutional Networks Convolution operation Motivation- Pooling Variants of the Basic Convolution function
CNN Architecture Understanding and Visualizing CNN LeNet AlexNet VGG.
Unit III SEQUENCE MODELLING 9
Recurrent and recursive nets: Recurrent neural networks Bidirectional RNN Encoder decoder sequence to
sequence architecture - Recursive neural networks introduction to Long term dependencies LSTM.
Unit IV AUTOENCODERS AND GENERATIVE ADVERSARIAL NETWORKS 9
Introduction to autoencoders Undercomplete autoencoders Regularized Autoencoders. Generative adversarial
networks Deep convolutional Generative adversarial networks (DCGAN) SRGAN CycleGAN.
Unit V APPLICATIONS 9
Computer vision: Object detection Localization Recognition - Natural Language Processing Neural language
models Neural machine translation using attention mechanism and aligning pieces of data.
Reference Book:
1. YoshuaBengio and Ian J.Goodfellow and Aaron Courville, "Deep Learning", MIT Press, 2016.
2.
3.
edition, 2019.
4.
5. -On Machine Learning with Scikit-Learn and TensorFlow: Concepts, Tools, and

Extensive Reading:
http://www.deeplearning.net
www.cs.toronto.edu/~fritz/absps/imagenet.pdf
http://neuralnetworksanddeeplearning.com/

PAGE 16
MEPCO R2023 ME COMPUTER SCIENCE AND ENGINEERING

23MC252 DEEP LEARNING LABORATORY L T P C

Offered by CSE (Core Practical) 0 0 3 1.5
Course Objective:
To impart knowledge on
To practice python packages for Deep Learning
To explore various CNN architecture.
To work with time series data using RNN and LSTM
To solve problems using Autoencoders and Generative Adversarial Neural Networks
To implement real world problems using Deep learning techniques
Course Outcome:
At the end of the course, the students will be able to
Implement Deep neural networks using Python libraries
Develop programs to perform image classification and object detection
Design and implement problems to use time series data using RNN and LSTM
Develop programs for Denoising images
Solve real world problems using Deep learning algorithms
Practical Course 45
Practical Syllabus:
It is recommended that all implementations are carried out in Deep Learning Framework preferably in Tensorflow.
1. Working with Tensorflow environment
2. Implementation Feed forward neural network with different activation functions and optimizations
3. Implementation of CNN architecture to recognize hand written data / OCR dataset
4. Implementation of Image classification algorithms using CNN
5. Implementation of object detection algorithms using CNN
6. Implementation of object detection in multiscene data using CNN
7. Implementation of various CNN architecture and analyse performances
8. Implementation of RNN and LSTM using time series data
9. Implementation of application to generate and clear noises present in images
10. Implementation of programs using various autoencoders and generative adversarial networks
11. Build neural language models
12. Build machine translation models
Reference Book:
1. YoshuaBengio and Ian J.Goodfellow and Aaron Courville, "Deep Learning", MIT Press, 2016.
2. Vinita
3.
edition, 2019.
4.
5. -On Machine Learning with Scikit-Learn and TensorFlow: Concepts, Tools, and

Extensive Reading:
http://www.deeplearning.net
www.cs.toronto.edu/~fritz/absps/imagenet.pdf
http://neuralnetworksanddeeplearning.com/
NPTEL, courseera course materials

PAGE 19