U18AI605 DEEP LEARNING
Class: B.Tech. VI- Semester Branch: Computer Science and Engineering (AI & ML)
Teaching Scheme: Examination Scheme:
L T P C Continuous Internal Evaluation 40 Marks
3 - - 3 End Semester Exam 60 Marks
Course Learning Objectives (LOs):
This course will develop student’s knowledge in/on…
LO1: deep learning fundamental and regularization parameters
LO2: principles and architectures of convolutional neural networks
LO3: principles and architectures of recurrent and recursive neural networks
LO4: practical methodology and applications of deep neural networks
UNIT – I (9)
Introduction: Why Deep Learning? Deep learning architecture, Historical trends in deep
learning and its performance, deep feed forward networks, Gradient based learning, Hidden
Units, Back propagations and differentiation algorithms, Case study: Deep learning applications
and societal impact
Regularization for Deep learning: Parameters norm penalties, Norm penalties as constrained
optimization, Regularization and under constrained problems, Dataset augmentation, Noise
robustness, semi-supervised learning, multitask learning, Parameter tying and sharing, Sparse
representation
UNIT –II (9)
Optimization for Training Deep Models: Pure optimization, challenges in neural network
optimization, Basic algorithms, Parameter initialization strategies, Algorithms with adaptive
learning rates
Convolutional Neural Networks: Convolution operations, Motivation, Pooling, Convolution
and pooling as an infinitely strong prior, Variants of the basic convolution functions, Structured
outputs, Data types, Efficient convolution algorithms, random or unsupervised features,
Convolution networks and the history of deep learning TensorFlow case study: Using
Convolutional Neural Networks for recognizing and classifying images
UNIT– III (9)
Recurrent Neural Networks: Architecture, Unfolding computational graphs, Recurrent neural
networks, Bidirectional Recurrent neural networks, Encoder-decoder architectures, Deep
recurrent networks
Recursive Neural networks: Architecture, Challenges of long-term dependencies, Echo state
networks, Leaky units, Strategies for multiple time scales, Long and short term memory for
recursive neural networks, Gated recursive neural networks, optimization for long term
dependencies, Explicit memory
Case study: Tensor flow modeling for recognition of hand-written digits
UNIT - IV (9)
Practical Methodology: Performance Metrics, Default baseline models, selecting hyper
parameters, debugging strategies, Case study: Multi digit number recognition
Applications: Large scale deep learning, Computer vision using deep learning, Speech
recognition using deep learning, Natural language processing using deep learning, other deep
learning applications,
TensorFlow case studies: Textual document processing, Deep learning on cloud
Text Book(s):
[1] Ian Goodfellow and Yoshua Bengio and Aaron Courville, Deep Learning, 1st ed., MIT Press,
2017.
[2] Antonio Gulli, Amita Kapoor, Sujit Pal, Deep Learning with TensorFlow 2.0 and Keras, 2nd ed.,
Mumbai: Packt, 2019 (Chapters: 1, 3, 4, 5).
Reference Books:
[1] Deng & Yu, Deep Learning: Methods and Applications, 1st ed., USA: Now Publisher, 2013
[2] Ian Goodfellow, Yoshua Bengio, Aaron Courville, Deep Learning, 1st ed., Massachusetts: MIT
Press, 2016.
[3] Sebastian Raschka, Vahid Mirjalili - Python Machine Learning, 2nd ed., Mumbai: Packt, 2019.
Course Research Paper: Research paper (indexed Journal/conference paper) relevant to the
course content will be posted by the course faculty in CourseWeb page.
Course Patent: Patent relevant to the course content will be posted by the course faculty in
CourseWeb page.
Course Projects: Course project is an independent project carried out by the student during the
course period, under the supervision of course faculty. Course faculty will post few course
project titles in CourseWeb page. Students are encouraged to come up and experiment with the
ideas that interest them.
Course Learning Outcomes (COs):
On completion of this course, students will be able to ...
CO1: apply deep learning fundamentals and regularization techniques for tuning the performance of the deep
learning models
CO2: design and analyze deep convolutional neural network models for image recognition
CO3: apply recurrent and recursive neural network architectures for handling long term dependencies and
vanishing gradient problem
CO4: design deep learning models for computer vision, speech and natural language processing applications
Course Articulation Matrix (CAM): U18AI605 DEEP LEARNING
Course Outcomes PO1 PO2 PO3 PO4 PO5 PO6 PO7 PO8 PO9 PO10 PO11 PO12 PSO1 PSO2 PSO3
CO1 U18AI605.1 2 2 2 1 - 1 - 1 1 1 - 1 2 1 2
CO2 U18AI605.2 2 2 1 2 2 - - 1 1 1 - 1 2 1 2
CO3 U18AI605.3 2 2 1 3 2 - - 1 1 1 - 1 2 1 2
CO4 U18AI605.4 - 2 2 3 2 - - 1 1 1 - 1 2 1 2
U18AI605 2 2 1.5 2.25 2 1 - 1 1 1 - 1 2 1 2
Internal Members External Members
Dr. S. Narasimha Reddy Prof. G. Sanjeevi
Dr. V. Shankar Dr. P. Sateesh Kumar
Dr. A. Jothi Prabha Smt. K. Sravanthi
Dr. V. Swathy Sri. P. Prasad
Sri. B. Srinivas Sri. K. Raja Gopal
Sri. K. Venkata Swamy
Dr. Praveen Jambholkar
Sri Koushik Bhupathi
