SEMESTER V
Course Course Course T L I C
V20UDS501 Deep Learning C Professional Core Course
Code Name Category 5
Pre-requisite Courses Nil Co-requisite Courses Nil Progressive Courses Nil
Course Offering Department Computer Applications Data Book / Codes/Standards Nil
Course Learning Rationale (CLR): The purpose of learning this course is to: Learning Program Learning Outcomes (PLO)
CLR-1 : Understand the concepts of Deep Learning & Neural Networks 1 2 3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
CLR-2 : Understand Deep neural network and layered learning approach
CLR-3 : Study and understand CNN and RNN for deep learning
Link with Related Disciplines
Ability to Utilize Knowledge
CLR-4 : Learn and understand Auto Encoders and its applications
Level of Thinking (Bloom)
Expected Proficiency (%)
Fundamental Knowledge
Expected Attainment (%)
Application of Concepts
Analyze, Interpret Data
CLR-5 : Understand concept of transfer learning and its applications with keras
Procedural Knowledge
Problem Solving Skills
Communication Skills
Skills in Specialization
Professional Behavior
To learn algorithmic topics of machine learning and mathematically deep enough to introduce the
Life Long Learning
Investigative Skills
Skills in Modeling
CLR-6 : required theory.
Analytical Skills
ICT Skills
Course Learning Outcomes (CLO): At the end of this course, learners will be able to:
CLO-1 : Apply basic mathematical concepts in Deep Learning 2 80 70 H H H H H M L M H M - H H H M
CLO-2 : Work with powerful framework for supervised learning 3 85 75 H H H H H M L M H M - H H H M
CLO-3 : Deal with Convolution Neural Networks 3 75 70 H H H H H M L M H M - H H H M
CLO-4 : Analyze various types efficient data encoders 3 85 80 H H H H H M L M H M - H H H M
CLO-5 : Apply various network models in deep learning 3 85 75 H H H H H M L M H M - H H H M
CLO-6 : Apply basic mathematical concepts in Deep Learning 3 80 70 H H H H H M L M H M - H H H M
Duration
18 18 18 18 18
(hour)
Introduction to deep learning
SLO-1 Deep feed forward networks Convolutional Networks Auto encoders Attention Models
S-1
Neural Networks Basics Introduction to Attention Models in
SLO-2 Learning XOR Convolutional operation Decoder
Vision
Various paradigms of earning Variational auto-encoders
SLO-1 Gradient-Based Learning Motivation Vision and Language
problems
S-2
SLO-2 Learning Algorithms Logistic Regression Cost Function Convolutional layers Generative adversarial networks Image Captioning
Multilayer perceptron–The first Additional layers
Supervised and Unsupervised
S-3 SLO-1 example of network with Keras Auto encoder VS PCA I Visual QA
Training
code
55
Residual Nets
Linear Algebra for machine Visual Dialog
SLO-2 error functions Auto encoder VS PCA II
Learning, Dimensionality reduction
Deep Generative Models
SLO-1 Cross-Validation Nested Functions Normalization Auto encoder Training I
S-4
SLO-2 Dimensionality reduction Architecture Design Pooling Auto encoder Training II GANs, VAEs
SLO-1 Lab 4 - Write a program to Lab 7 –Training CNN
generate XOR function using Examples from Computer Vision:
S Lab 1 - Building blocks of deep Lab 10 – Auto encoders with Keras, Lab 13 - Image Editing, In painting,
McCulloch-Pitts neuron and
5-6 SLO-2 neural networks. TensorFlow, and Deep Learning Super resolution
appropriate values for Classification examples (AlexNet)
Weights, bias and threshold.
SLO-1 Over/Under-fitting differentiation algorithms Applications in Computer Vision Under complete Alexnet to ResNet
S-7
Hyper parameters and validation
SLO-2 Regularization for Deep learning Image Net Regularized Transfer learning
sets
SLO-1 Estimators Early Stopping Generative adversarial neural stochastic Siamese Networks
S-8 networks
Bias
SLO-2 Drop out Strided Convolutions Denoising Metric Learning
SLO-1 Variance Unit saturation Convolutions Over Volume Contractive Ranking/Triplet loss
S-9 Heuristics for avoiding bad local One Layer of a Convolutional
SLO-2 Regularization Applications RCNNs
minima Network
SLO-1 Introduction to a simple DNN Heuristics for faster training Inception Network Motivation Dimensionality reduction CNN-RNN
S-10 Nestors accelerated gradient
SLO-2 Platform for deep learning Transfer Learning Classification PixelRNNs
descent
SLO-1 Lab 2 - Understand the key
S computations underlying deep Lab 5 - Write a program for
Lab 8 –Work with Deep Lab 11 - Denoisingautoenecoders Lab 14 - 3D Object Generation
11- learning, use them to build and train perceptron net for an AND function
SLO-2 convolutional models with Keras and TensorFlow Few-shot Learning
12 deep neural networks, and apply it with bipolar inputs and targets.
to computer vision.
SLO-1 Dimensionality reduction Greedy layer wise training Sequence Modeling Recommendation CycleGANs
S-13 Optimization for Deep Learning-
SLO-2 Deep learning software libraries Regularization VGGNet, LeNet Progressive GANs
Optimizers
Perspectives and Issues in deep Explanation of logistic regression
SLO-1 Sequential model RMS Prop for RNNs StackGANs
learning cost function
S-14
Softmax cross entropy loss
SLO-2 Back propagation Recurrent Neural Networks SGD for CNNs Pix2Pix
function.
simple examples and motivation for Applications in captioning and video
SLO-1 Training Example Networks RNN topologies Momentum Optimizer
S-15 deep network tasks
SLO-2 Shallow neural networks Stochastic Gradient Descent Difficulty in Training RNN NAG optimizer 3D CNNs
56
Risk minimization, loss function,
SLO-1 ADAM Adagrad Optimizer Zero-shot
back propagation Long Short Term Memory
S-16 Tensor flow
review of fundamental learning
SLO-2 Bidirectional LSTMs Adadelta and Adam Optimizer One -shot
techniques
SLO-1 Lab 6 Write a program (with a
suitable example) to demonstrate
Lab 3 - Learn to build a neural how the hyper plane is changing in
S Lab 9 – Application case study – Lab 12 - Anomaly detection with Lab 15 - Self-supervised Learning;
network with one hidden layer, different
17- Hand written digits recognition using Keras, TensorFlow, and Deep Reinforcement Learning in Vision
SLO-2 using forward propagation and back Iterations using the perceptron
18 deep learning, LSTM with Keras Learning
propagation learning law with its decision
regions. Give the output in
graphical form.
Li Deng, Dong Yu, “Deep Learning: Methods and Applications”, Foundations and Trends in Signal Processing
Michael Nielsen, “Neural Networks and Deep Learning”, Online book, 2016 ([Link]
Learning
Ian Goodfellow, YoshuaBengio, Aaron Courville, “Deep Learning”, MIT Press, 2016 (available at [Link]
Resources
Kevin P. Murphy, “Machine Learning: A Probabilistic Perspective”, MIT Press, 2012
Jason Brownlee , “Deep Learning with Python” , ebook, 2016
Goodfellow, I., Bengio,Y., and Courville, A., Deep Learning, MIT Press, 2016..
Course Designers
Experts from Industry Experts from Higher Technical Institutions Internal Experts
[Link]
57