Announcements Deep Learning Foundations and Applications Quick Links
Coming up in an OBE format!

*OBE is outcome-based education according to Washington Accord

13 Feb. 2020 Download Old Question Papers.

27 Dec. 2019 Classes start on Mon, 6 Jan. 2020.

Open for UG3, UG4, PG1 and MS/PhD only. Limited to 200 seats, apply through ERP within 2 Jan 2020. Shortlisting would be declared by 3 Jan 2020.

UG/PG/DualDegree students to apply on ERP and await approvals.
MS/PhD to apply directly through ERP as Recommended or Additional coursework subject only.
As a matter of inclusiveness and fairness policy with seat limitations we are not able to acknowledge audit requests.
Students are expected to be proficient in Python programming.
Spring 2020

Subject Type: Elective | LTP: 3-1-0 | Credits: 4
Location: NR222, Nalanda Lecture Hall Complex, IIT Kharagpur
Time: Slot C / Mon (10:00 AM - 10:55 AM) + Wed (8:00 AM - 9:55 AM) + Thu (10:00 AM - 10:55 AM)

Instructors: Dr. Debdoot Sheet, Prof. Sudeshna Sarkar
TAs: Rachana Sathish

Grading: Attendance 10%, Quizzes 10%, Coding Assignments 20%, Mid-Term 25%, End-Term 35%
Assignment Submission through Moodle. (How to bypass proxy server for access)
Linear Algebra - Gilbert Strang
A Gentle Introduction to Programming Using Python - Sarina Canelake
Design and Analysis of Algorithms - Dana Moshkovitz and Bruce Tidor

Tools of the Trade: Anaconda Python 3.6 | PyTorch | Getting started with PyTorch

Related MOOCs: Deep Learning for Visual Computing | Github repository

Previous years: Spring 2019

Why this subject?
This subject aims to provide students with foundational concepts required for deep learning which is now prevalent across various applications ranging across speech and natural language processing to machine vision to medical imaging. The course will introduce the fundamental principles of deep neural networks and the important paradigms of deep learning. On having studied this subject a student is expected to be able to build analytics solutions to problems in signal, image and text paradigm using deep neural networks.

The course will focus on demonstrating different applications of deep neural networks to a number of examples across domains including computer vision and NLP. The course will contain tutorials which will focus on hands-on session and implementation of deep neural networks and applications that use them. Students on completion are expected to be able to understand the concepts of deep neural networks and will be able to develop solutions using deep neural networks.

Text books:
[1]. I. Goodfellow, Y, Bengio, A. Courville, Deep Learning, MIT Press, 2016.
[2]. S. Haykin, Neural Networks and Learning Machines, 3rd Edition, Pearson, 2008.

Reference books:
[R1]. C. M. Bishop, Neural Networks for Pattern Recognition, Oxford University Press, 1995.
[R2]. R. O. Duda, P. E. Hart, D. G. Stork, Pattern Classification, 2nd Edition, Wiley, 2001.
[R3]. D. Cohen-Or, C. Greif, T. Ju, N. J. Mitra, A. Shamir, O. Sorkine-Hornung, H. Zhang, A Sampler of Useful Computational Tools for Applied Geometry, Computer Graphics and Image Processing, CRC Press, 2015.
[R4]. T. M. Mitchell, Machine Learning, Mc. Graw Hill Education, 1997.
[R5]. C.M. Bishop, Pattern Recognition and Machine Learning, 2nd Edition, Springer, 2011.
[R6]. S. Russel and P. Norvig, Artificial Intelligence: A Modern Approach, 3rd Edition, Prentice Hall/Pearson, 2015.

Measure of Outcome:
A student undertaking this subject would be graded based on perfromance in all of the following:
(1) Regular participation in class activity.
(2) Timely submission of all quizzes, online assignments.
(3) Participation in tutorials in class.
(4) Appear for all the exams.
(5) Also attend the practice tutorials and workshops.
Foundational concepts of linear algebra, probability, neural networks and learning theory
Familiarity with software toolkits and deep learning libraries
Convolutional and recurrent neural networks
Regularization and learning concepts
Basics of dataset handling for deep learning
Numerical precision of deep learning computation
Quantifying perception losses and Adversarial learning
Federated deep learning
Deep learning for machine translations and text summarization
Building chatbots with cognitive capability
Deep learning for image classification, scene understanding
Semantic segmentation and single shot multibox detection
Medical image classification, compression and super resolution