R22 BToch. CSE (Al and ML) Syllabus UNTU Hyderabad AM701PC: DEEP LEARNING B.Tech. IV Year! Sem. eq ov eo Course Objectives: + To.understand deep Learning algorithms and their applications in real-world data Course Outcome: ‘© Understand machine leaming basics and neural networks ‘© Understand optimal usage of data for training deep models + Apply CNN and RNN models for real-world data + Evaluate doop models © Develop deep models for real-world problems UNIT-1 Machine Learning Basics Learning Algorithms, Capacity, Overfiting and Underfiting, Hyperparameters and Validation Sets, Estimators, Bias and Variance, Maximum Likelihood Estimation, Bayosian Statistics, Suporvisod Loaming Algorithms, Unsupervised Leaming Algorithms, Stochastic Gradiont Descent, Building a Machine Learning Algorithm, Challenges Motivating Deep Learning Deep Feedforward Networks Leaming XOR, Gradiant-Based Learning, Hidden Units, Architecture Design, Back-Propagation and Qthor Differontiation Algorithms UNIT =I Regularization for Deep Learning Parameter Norm Penalties, Norm Penalties as Constrained Optimization, Regularization and Under Constrained Probloms, Dataset Augmentation, Noiso Robustness, Somi-Suporvised Learning, Multi Task Loarning, Early Stopping, Parameter Tying and Parameter Sharing, Sparse Representations, Bagging and Other Ensemble Methods, Dropout, Adversarial Training, Tangent Distance, Tangent Prop, and Manifold Tangent Classifier, Optimization for Training Deap Modals, Learning vs Pure Optimization, Challenges in Neural Network Optimization, Basic Algorithms, Parameter Initialization Stratagias, Algorithms with Adaptive Learning Rates UNIT =I Convolutional Networks ‘The Convolution Operation, Motivation, Pooling, Convolution and Pooling as an Infinitely Strong Prior, Variants of the Basic Gonvolution Function, Structured Outputs, Data Types, Efficient Convolution Algorithms, Random or Unsupervised Featuros UNIT-1V Recurrent and Recursive Nets Unfolding Computational Graphs, Recurrent Neural Networks, Bidirectional RNNs, Encoder-Decoder Sequence-to-Sequence Architectures, Deep Recurrent Networks, Recursive Neural Networks, The Challongo of Long-Term Depondoncios, Echo Stato Networks, Leaky Units and Other Stratogios for Multiple Time Scales, The Long Short-Term Memory and Other Gated RNNs, Optimization for Long- Torm Dependencies, Explicit Mamory UNIT-Vv Practical Methodology: Performance Metrics, Default Baseline Models, Determining Whether to Gathor Moro Data, Solocting Hyperparamotors, Debugging Strategios, Example: Mult:Digit Number Recognition Page 106 of 147R22 BToch. CSE (Al and ML) Syllabus UNTU Hyderabad Applications: Large-Scale Deep Learning, Computer Vision, Speech Recognition, Natural Language Processing, Other Applications. TEXT BOOK: 1 Desp Learning by lan Goodfellow, Yoshua Bongio and Aaron Courvill, MIT Pross. REFERENCE BOOKS: 1 ‘Tho Eloments of Statistical Learning. Hastio, R. Tibshirani, and J. Fridman, Springer. Probabilistic Graphical Models. Kollor, and N. Friedman, MIT Pross. Bishop. C.M., Pattern Recognition and Machine Learning, Springor, 2006, Yegnanarayana, B., Arlificial Noural Networks PHI Learning Pvt. Ltd, 2009. Golub, G.H., and Van Loan, C.F, Matrix Computations, JHU Press, 2013. Satish Kumar, Neural Networks: A Classroom Approach, Tata McGraw-Hill Education, 2004, Page 107 of 147