IRMA UNIVERSITY Institute: Institute of Technology ‘Name of Programme: | MTech CSE (Cyber Security) ‘Course Cod 6CS401 ‘Course Tith Machine and Deep Learning | ‘Course Type: ‘(Corel “| Value Added Course /(1 Department Elective / 1 Institute 1 University Elective/ © Open Elective / Any other) ‘Year of Introduction: [2022-23 L_|T | Practical Component | C Lew | Pw [Ww [Ss 3[-[ 2 -[-[-[4 Course Learning Outcomes (CLO): At the end of the course, the student will be able to — 1. make use of machine leaming and deep lea problems in applicable domains ig techniques to solve (BL3) 2. appraise the need of machine learning and deep learning (BLS) 3. evaluate variety of machine leaming algorithms for appropriate (BLS) applications. 4, adopt different deep learning algorithms appropriate to solve problems in (BL6) various domains Syllabus: Total Teaching hours: 45 Syllabus Teaching hours Unit Introduction: Motivation and Applications, importance of Data 02 Visualization, Basics of Supervised and Unsupervised Learning, Unsupervised Learning 03 Clustering: Hierarchical Agglomerative Clustering, k-means Algorithm, Self-Organizing Maps Unit-III Supervised Learning 12 Regression Techniques: Basic concepts and applications of Regression, Simple Linear Regression — Gradient Descent Method, Multiple Linear Regression, Non-Linear Regression, Linear Regression with Regularization, Hyper-parameters tuning, Loss Functions, Evaluation Measures for Regression Techniques Classification Techniques: Naive Bayes Classification, Fitting Multivariate Bemoulli Distribution, Gaussian Distribution and Multinomial Distribution, K=Nearest Neighbours, Decision trees. Support Vector Machines: Hard Margin and Soft Margin, Kernels and Kernel Trick, Evaluation Measures for Classification Techniques UnitIV Artificial Neural Networks: Perceptron Leaming, Feed Forward 10 ‘Neural Networks, Back-propagation, Unstable and vanishing Gradient Problem deUnit-V Unit-VI Unit-VIT Convolutional Neural Networks: Convolution & Pooling, Dropout, Batch Normalization, State-of-the-art CNNs Transfer Learning & Domain Adaptation: Transfer Learning Scenarios, Applications of Transfer Leaming, Transfer Leaning Methods, Fine Tuning and Data Augmentation, Supervised, Semi Supervised and Unsupervised Deep Learning Sequence Models: Recurrent Neural Networks (RNN), Language Modelling, Long-Short Term Memory Network, Gated Recurrent Unit, Biedirectional RNN, Applications of Sequence Models Miscellaneous: Linear Discriminant Analysis, Auto encoders and Stacked Auto encoders, Generative Adversarial Networks, Deep 03 09. 06 Reinforcement Learning Self-Study: Suggested Readings! References: Suggested List of Experiments: The self-study contents will be declared at the commencement of semester. Around 10% of the questions will be asked from self-study contents 1, Tom Mitchell, Machine Learning, TMH 2. C. Bishop, Patter Recognition and Machine Learning, Springer 3. Kishan Mehrotra, Chilukuri Mohan and Sanjay Ranka, Elements of Artificial Neural Networks, Penram International 4, Athem Ealpaydin, Introduction to Machine Learning, PHI 5. Ian Goodfellow, Yoshua Bengio, Aaron Courville, Deep Learning, MIT Press, 6. Charu C. Aggarwal, Neural Networks and Deep Learning — A Textbook, Springer 7. Adam Gibson, Josh Patterson, Deep Learning, O'Reilly Media, Ine. 8. Hastie, T., Tibshirani, R. and Friedman, J. The Elements of Statistical Leatning, Springer sr. Title Hours No. 1 Introduction to Python, Numpy, Sklearn and 02 Matplottib. 2 Implementation of K-means and Hierarchical 02 clustering algorithms. 3 Linear Regression with Regularization (without using 02 skleatn or equivalent library) and Simple and Multiple Linear Regression with and without regularization using Skleam. 4 Nalve-Bayes — Multivariate Bernoulli, Multinomial 02 and Gaussian using sklearn. 5 Decision Trees — 1D3, C4.5 using sklear, Support 04 ‘Vector Classification and Regression with Grid Search for Hyper-parameter tuning using skleam, 6 AND gate using Perceptron Learning (self 04 implementation). Ex-OR Gate/any other problem*Marked as addi using Backpropagation Neural Networks (self implementation) and Sklearn. Convolutional Neural Network on MNIST, Fashion MNIST and CIFARIO datasets. Write code using (a) Sequential Class (b) Model Class API Use transfer learning for Image Segmentation & Detection using Deep Networks. Unsupervised MNIST Build a language model using RNN. Write functions to sample novel sentences and find the probability of input sentence. Also, use Recurrent Neural Network for Sentiment Analysis. Reinforcement learning for game development. jonal definitions for practice 04 04 04 02 06