Department of Computer Science, CUSAT

Home > Academic/Programmes > Programme Structure > CSE (2019)

19-475-0401: DEEP LEARNING ARCHITECTURES

Core/Elective: Core Semester: 4 Credits: 4

Course Description

Deep learning is part of a broader family of machine learning methods based on learning data representations, as opposed to task-specific algorithms. This course describes deep learning techniques used by practitioners in industry, including deep feed forward networks, regularization, optimization algorithms, convolutional networks, sequence modeling, and practical methodology. This course is useful to students planning careers in either industry or research, and for software engineers who want to begin using deep learning in their products or platforms

Course Objectives

To develop a clear understanding of the motivation for deep learning
To get a practical understanding of machine learning methods based on learning data
To design intelligent systems that learn from complex and/or large-scale datasets
To apply deep learning to practical problems

Course Content

Module I
Deep Networks: Feed forward networks – Learning XOR- Gradient based Learning – Hidden units – Architecture design- Back propagation – Differentiation algorithms

Module II
Regularization for Deep Learning: Penalties-Constrained optimization-Under constrained problems- Dataset augmentation-Semi Supervised learning- Sparse representation- Adversarial training- Optimization for training deep models: Basic algorithms-Algorithms with adaptive learning rates

Module III
Convolutional Networks: Convolution-Pooling-Variants of pooling- Efficient convolutional algorithms – Recurrent and Recursive Nets: Recurrent Neural Networks-Deep Recurrent Networks-Recursive Neural Networks- Explicit memory

Module IV
Linear Factor Models: Probabilistic PCA- ICA – Slow feature analysis – Sparse coding – Autoencoders: UndercompleteAutoencoders – Regularized Autoencoders- Learning Manifolds-Applications of Autoencoders – Representation learning

Module V
Deep generative models: Boltzmann Machines – RBM- Deep Belief Networks-Deep Boltzmann MachinesConvolutional Boltzmann Machines- Directed generative Nets

REFERNCES

1. Ian Goodfellow, YoshuaBengo, Aaron Courville, Deep Learning, 1e, MIT Press, 2017
2. Nikhil Buduma and Nicholas Locascio, Fundamentals of Deep Learning: Designing Next-Generation Machine Intelligence Algorithms, 1e, Shroff/O'Reilly, 2017
3. Josh Patterson and Adam Gibson, Deep Learning: A Practitioner's Approach, 1e, Shroff/O'Reilly, 2017