This course covers theories and practices of Artificial Intelligence. Under this course we study about intelligent agents which can receive percepts from the environment and perform intelligent actions. Each such agent implements a function that maps percept sequences to actions. In this course, we will learn how such an agent formalizes its goal, and what algorithms it uses to achieve those goals by performing a sequence of optimal actions. The topics covered in this course are decomposed in the following four parts: i) problem-solving, ii) knowledge representation, and inference, iii) incorporating uncertainty in reasoning methodologies, and iv) learning. These topics will help us to learn the followings: how to represent knowledge, how to generate and evaluate appropriate sequence of actions, how to search among the choice of actions to obtain optimal or near-optimal solution, and how to incorporate uncertainty and how to learn from experience.
The course is an introduction to deep learning. Under this course we will study about deep learning models, and algorithms. In the first two weeks of the course, a fast-paced review of machine learning, and gradient descent based optimization will be provided. Then we will cover traditional deep learning topics, such as logistic regression, feed forward neural networks, autoencoders, convolution neural networks, recurrent neural networks, LSTM, graph neural networks, transformers, deep generative model, adversarial and reinforcement learning. Towards the end of the semester, we will study selected advanced topics in deep learning that have appeared in recent machine learning conferences. Student should have completed an undergraduate machine learning or data mining course as pre-requisite.