18CS3059 Machine Learning Course Objectives: Enable the student to 1. understand the existing machine learning techniques: it’s concepts, mathematical background, applicability, limitations. 2. design and analyze various machine learning based applications with a modern outlook focusing on recent advances. 3. evaluate the performance of the models developed Course Outcomes: The student will be able to 1. describe some concepts and methods central to machine learning such as classification, regression, clustering, bias/variance. 2. explain mathematically various machine learning approaches and paradigms. 3. compare the strengths and limitations of selected machine learning algorithms and where they can be applied in different applications. 4. design and implement suitable machine learning algorithm to a given task 5. apply some state-of-the-art development frameworks and software libraries in machine learning task realization. 6. evaluate the performance of machine learning algorithms using suitable metrics Module 1: Introduction Learning – Types of Machine Learning – Supervised Learning – The Brain and the Neuron – Design a Learning System – Perspectives and Issues in Machine Learning – Concept Learning Task – Concept Learning as Search – Finding a Maximally Specific Hypothesis – Version Spaces and the Candidate Elimination Algorithm. Module 2: Supervised Learning (Linear Models) Basic methods: Distance-based methods, NearestNeighbors, Decision Trees, Naive Bayes, Linear Regression, Logistic Regression, Generalized Linear Models, Perceptron. Module 3: Supervised Learning (Non-Linear Models) and Ensemble Learning Multilayer Perceptron, Radial Basis Function, Support Vector Machines, Nonlinearity and Kernel Methods, Ensemble Learning: Boosting, Bagging, Random Forests. Module 4: Unsupervised Learning Unsupervised Learning, K means Algorithms, Vector Quantization, Self-Organizing Feature Map, Partitioning Methods, Hierarchical Methods, Density based Methods, Grid based Methods, Model based Clustering Methods, Clustering High-Dimensional Data. Module 5: Dimensionality Reduction Dimensionality Reduction: Principal Component Analysis, Linear Discriminant Analysis, Factor Analysis, Independent Component Analysis, Locally Linear Embedding, Canonical Correlation Analysis, Isomap Module 6: Probabilistic Learning and Reinforcement Learning Introduction to Probabilistic Learning, Gaussian Mixture Models, EM Algorithm, Nearest Neighbour Methods, Elements of Reinforcement Learning, Model based Learning, Temporal Difference Learning, Generalization. Reference Books: 1. Ethem Alpaydin, “Introduction to Machine Learning”, (Adaptive Computation and Machine Learning Series), Third Edition, MIT Press, 2014, ISBN-10: 0262028182, ISBN-13: 978- 0262028189 2. Stephen Marsland, “Machine Learning – An Algorithmic Perspective”, Second Edition, Chapman and Hall/Crc Machine Learning and Pattern Recognition Series, 2014, SBN-10: 1466583282, ISBN-13: 978-1466583283 3. Kevin Murphy, “Machine Learning: A Probabilistic Perspective”, MIT Press, 2012, ISBN10: 0262018020, ISBN-13: 978-0262018029 4. Christopher Bishop, “Pattern Recognition and Machine Learning” (Information Science and Statistics), Springer, 2007. 5. Trevor Hastie, Robert Tibshirani, Jerome Friedman, “The Elements of Statistical Learning: Data Mining, Inference, and Prediction”, Second Edition (Springer Series in Statistics), 2016, ISBN10: 0387848576, ISBN-13: 978-0387848570 6. Y. S. Abu-Mostafa, M. Magdon-Ismail, and H.-T. Lin, “Learning from Data”, AMLBook Publishers, 2012 ISBN 13: 978-1600490064. 7. P. Flach, “Machine Learning: The art and science of algorithms that make sense of data”, Cambridge University Press, 2012, ISBN-10: 1107422221, ISBN-13: 978-1107422223 17CS3032 APPLIED MEDICAL IMAGE PROCESSING Course Objectives: Enable the student to • understand the basics of medical image datasets and the special requirements for processing medical imaging data. • recognize the need of image processing techniques for clinical applications. • apply various imaging techniques like transformations, rendering, registration, and reconstruction on medical images. Course Outcomes: The student will be able to • describe the specific terminologies and objectives of utilizing different imaging modalities in clinical practice . • infer the intensity and filtering operations on medical image datasets. • summarize the segmentation and multimodal image registration techniques used for clinical applications. • analyze the application of imaging techniques like transformations and rendering.. • explain the techniques and concepts of image registration and reconstruction. • predict the issues in medical image archiving, retrieval and communication and apply the techniques to real world problems. Unit I - Basics of medical image sources: Radiology-Electromagnetic spectrum-Attenuation and imaging - CT - MRI - Ultrasound- Nuclear medicine and molecular imaging-Other imaging techniquesRadiation protection and dosimeter- Image representation-Pixels and voxels-Gray scale and color representation- Image file formats DICOM-Other formats-Image quality and Signal-noise ratio. Unit II - Operations in intensity space: Intensity transform function and dynamic range-WindowingHistograms and Histogram operations-Dithering and depth- Filtering and transformation-Filtering operation- Fourier Transform Other transforms- Segmentation -ROI definition and centroidsThresholding- Region Growing- Segmentation methods- Morphological operations-Evaluation of segmentation results Unit III - Spatial transforms: Discretization-Resolution and Artifacts-Interpolation and Volume regularization Translation and Rotation- Reformatting-Tracking and Image guided therapyRendering and Surface models Visualization-Orthogonal and Perspective projection and the viewpoint-Ray casting- Surface based rendering Unit IV - Image registration: Fusing information- Registration paradigms- Merit functionsOptimization strategies-Camera calibration- Registration to physical space-Evaluation of registration results- CT reconstruction Radon transform- Algebraic reconstruction-Filtered back propagation Unit V - Image compression: Fundamentals and standards of compression and communicationMedical image archive, retrieval and communication- Quality evaluation for compressed medical images- Three dimensional image compression with wavelet transforms. Reference books: 1. Wolfgang Birkfellner, Applied Medical Image Processing: A Basic Course, CRC Press, 2011, ISBN: 1439824452, 9781439824450 2. Isaac Bankman, Handbook of Medical Image Processing and Analysis, Second edition, Academic Press, 2008, ISBN: 008055914X, 9780080559148 3. Geoff Dougherty, Digital Image Processing for Medical Applications, Cambridge University Press, 2009, ISBN: 0521860857, 4. Geoff Dougherty, Medical Image Processing: Techniques and Applications, Springer, 2011, ISBN: 1441997792, 9781441997791 5. Paul Suetens, Fundamentals of Medical Imaging, Cambridge University Press, 2009, ISBN: 0521519152, 9780521519151 1 18CS3106 Deep Learning Course Objectives: Enable the student to 1. study the basic concepts of neural networks and deep learning 2. comprehend deep learning techniques 3. explore various applications for deep learning techniques Course Outcomes: The student will be able to 1. understand the basics of deep learning 2. implement various deep learning models 3. realign high dimensional data using reduction techniques 4. analyze optimization and generalization in deep learning 5. explore the deep learning applications 6. apply the algorithms to real time problems Module 1:Machine Learning Machine learning- Linear models (SVMs and Perceptron’s, logistic regression) - Intro to Neural Networks - Training a neural network: loss functions, backpropagation and stochastic gradient descent - Neural networks as universal function approximates Module 2: Deep Neural Networks Deep Learning- A Probabilistic Theory of Deep Learning- Deep Forward Networks - Backpropagation and regularization, batch normalization- VC Dimension and Neural Nets-Deep Vs Shallow Networks Module 3: Convolutional Neural Networks Convolutional Neural Network - Architectures - AlexNet, VGG, Inception, ResNet - Training a Convnet: weights initialization, batch normalization, hyperparameter optimization Module 4: Optimization Optimization in deep learning - Non-convex optimization for deep networks- Stochastic Optimization Generalization in neural networks - Vanishing gradient problem and mitigation - Regularization - Dropout Module 5: Recurrent Neural Networks and Deep Unsupervised Learning Recurrent networks, LSTM, GRU - Architectures, Autoencoders and VariationalAutoencoders, Adversarial Generative Networks, DBM - Deep Reinforcement Learning Module 6: Applications Computer Vision- ImageNet- Detection- Face Recognition- Scene Understanding- Gathering Image Captions - Audio Wave Net - Natural Language Processing Word2Vec - Sentiment Analysis - Recent research Reference Books: 1. Ian Goodfellow, YoshuaBengio, Aaron Courville, “Deep Learning”, MIT Press, 2016. ISBN: 9780262035613 2. Deng & Yu, “Deep Learning: Methods and Applications”, Now Publishers, 2013. ISBN: 1601988141, 9781601988140 3. Michael Nielsen, “Neural Networks and Deep Learning”, Determination Press, 2015. 21CS3003 Artificial Intelligence and Machine Learning Course Objectives: Enable the student to 1. understand statistical and computational considerations in AI and ML methods. 2. apply the skill of devising computationally efficient and yet statistically rigorous algorithms for solving machine learning problems. 3. develop the skill of quantifying the statistical performance of any new AI and ML method. Course Outcomes The student will be able to 1. understand the existing machine learning techniques: it’s concepts, mathematical background, applicability, limitations and toolkit used in industries; 2. create AI/ML solutions for various societal problems 3. apply some state-of-the-art development frameworks and software libraries in machine learning task realization. 4. evaluate the performance of machine learning algorithms using suitable metrics. 5. compare the strengths and limitations of selected machine learning algorithms and where they can be applied in different applications. 6. build and deploy production grade AI/ML applications Module 1: Foundations of Machine Learning PAC Learning Framework– Empirical Risk Minimization (ERM), Uniform Convergence is Sufficient For Learnability; VC Theory– No Free Lunch Theorem, Fundamental Theorem of Learning; Non-Uniform Learnability– Structural Risk Minimization (SRM), Minimum Description Length and Occam’s Razor; Tools used in industries: Introduction and overview, Data Preprocessing, Feature Extraction algorithms. Module 2: Machine Learning Theory to Algorithms Linear Predictor and Boosting– Linear Regression and Logistic Regression, Boosting; Support Vector Machine– SVM, Kernel Methods; Decision Trees– Ensemble methods, Decision Tree Algorithms, Random Forests Clustering: K-Means, K Nearest Neighbours, Association Rule Learning; Dimensionality Reduction: PCA, SVD, Generalized Discriminant Analysis, Linear Discriminant Analysis, multidimensional scaling, Semi-supervised learning, Case Study (Clustering/Anomaly/Fraud Detection), Classical Reinforcement Learning, Markov Decision, Monte Carlo Prediction, Case Study (next best offer, dynamic pricing) Module 3: Foundations for Artificial Intelligence Problem spaces and search, state space graph, production systems BFS and DFS, Introduction to heuristic search, hill climbing, best first search, A* algorithm, admissibility, AND/OR graph – AO*, Predicate logic, rule‐based systems, forward vs backward reasoning, non‐monotonic reasoning, statistical reasoning, Dempster Shafer theory, Min‐Max search, Alpha‐Beta cut‐offs , Case studies: MYCIN, R1, Divide and Conquer, Greedy, Branch and Bound, Gradient Descent), NN basics (Perceptron and MLP, FFN, Back propagation) Module 4: Recurrent Neural Networks and Deep Learning Building recurrent NN, Long Short-Term Memory, Time Series Forecasting, Auto-encoders and unsupervised learning, Stacked auto-encoders and semi-supervised learning, Regularization Dropout and Batch normalization Module 5: Advanced Machine Learning concepts Cross-validation, Minimax Learning– ERM, Minimax Approach, Maximum Entropy Machine; Neural Networks/Deep Learning- CNN, RNN/LSTM/GRU, Transfer Learning, Case Study (CNN); Natural Language Processing: Text Mining, Generation (using case study), Case Study (Generation); Predictive Analytics – Forecasting, Logistic, Time Series (ARIMA),Case Study (Time Series), Module 6: Machine Learning and Artificial Intelligence Applications across Industries Recent trends in various application areas: Healthcare, Agriculture, Retail, Financial Services, Manufacturing, Hospitality etc. (using case studies) Text Books: 1. S. S. Shwartz and S. Ben-David, Understanding Machine Learning, Cambridge, 2016, ISBN13 : 978-1107512825 2. Kevin Knight, Elaine Rich and Shivashankar B. Nair, Artificial Intelligence, McGraw Hill Education; 3rd edition,2017, ISBN-10 : 9780070087705 Reference Books: 1. Y. S. Abu-Mostafa, M. Magdon-Ismail and H.-T. Lin, Learning From Data, AMLBook, 2012, ISBN-13 : 978-1600490064 2. N. J. Nilson, Principles of Artificial Intelligence, Narosa, 2002, ISBN-13: 978-8185198293. 3. P. Norvig, Paradigms of AI programming, 1st Edition, Elsevier, ISBN: 9781558601918 4. P. Jackson: Introduction to Expert System; Addison‐Wesley, Addison-Wesley Pub (Sd); 2nd edition, 1990, ISBN-13: 978-0201175783.