AIL2001 Introductory Data Science | 4 credits (3-0-2)
AI philosophy; data collection and models; data curation; data quality assessment; data infrastructure; data value; statistical training and hypothesis formation; statistical hypothesis testing; state spaces, local search and optimisation for machine learning; supervised and unsupervised machine learning; regression; decision trees and random forests; neural networks; hyperparameter tuning and performance metrics; model debugging; AI agents and sequential decision-making; application areas; AI/data ethics and social implications; overview of latest advances like large language models; practical assignments in python/R for familiarisation with data science and machine learning libraries, and interpretation of outcomes.
AIL7001 Mathematical Foundations of MINDS | 3 credits (3-0-0)
Vector space, dimension, basis, linear map, rank and nullity, system of linear equations, eigenvalues, diagonalization, inner product space, orthonormal basis, Gram Schimdt process, adjoint operator, SVD. Probability, Bayes theorem, pdf and cdf, some discrete and continuous distributions, moment generating function, concentration inequalities. Convex set, convex function and its properties, nonlinear programming, KKT optimality, Lagrange function, Lagrange dual, steepest descent method for unconstrained problem, stochastic gradient descent.
AIL7002 Data Structures and Algorithms | 4 credits (3-0-2)
Linear (arrays, linked lists, etc.) and non-linear data structures (trees, graphs, DAGs, etc.), hashing, sorting, and searching algorithms, divide-and-conquer approaches, dynamic programming, and greedy algorithms, suffix trees and inverted indices, complexity analysis, NP-completeness, approximation algorithms, linear and integer programming.
AIL7011 Numerical Optimization | 3 credits (3-0-0)
General properties of algorithm, global convergence, rate of convergence,unconstrained optimization techniques - one dimensional methods: Fibonacci method, Golden sectionmethod, interpolation. Multi-dimensional methods: pattern search method, Nelder and Meed method,gradient based methods: steepest descent method and its variants recall, conjugate direction andgradient method, quasi-Newton method. Constrained optimization techniques - penalty method,barrier method, projection gradient method, sequential quadratic programming approach. Semidefiniteand cone programming
AIL7013 Machine Learning and Optimization | 3 credits (3-0-0)
Convex sets and functions, gradient descent, stochastic gradient descent, and variants; constrained optimization (Lagrange multipliers, KKT conditions); duality; applications in training machine learning models; bias-variance tradeoff; empirical risk minimization; VC dimension and generalization bounds; structural risk minimization; regularization and its role in controlling model complexity; PAC learning framework; kernel methods and reproducing kernel Hilbert spaces (RKHS); Markov chains and Monte Carlo methods; introduction to deep learning theory; current trends.
AIL7014 Introduction to Optimization | 3 credits (3-0-0)
Optimization and loss functions, convergence of gradient descent, sub-gradient and conjugate gradient, aspects of SGD, constrained optimization andduality, non-convex optimization, online optimization, second-order methods, special topics (robust, distributed, federated) optimization.
AIL7021 Deep Learning | 4 credits (3-0-2)
Multilayer perceptron, shallow neural network, regression, classification, deep neural network, loss function, optimization, backpropagation, overfitting, regularization, convolutional neural networks, residual networks, RNN, LSTM, backpropagation through time (BPTT), sequence-to-sequence, attention mechanism, transformers, LLMs, graph neural network, deep reinforcement learning, generative adversarial networks, variational autoencoders, diffusion models, normalizing flows, self-supervised learning, contrastive learning, Bayesian neural networks, deep meta learning, deep continual learning.
AIL7022 Reinforcement Learning | 4 credits (3-0-2)
Markov decision processes, approximate dynamic programming, model-free prediction and control, Q-learning and SARSA, policy gradients, bandits, model-based RL, maximum entropy RL, imitation learning, inverse reinforcement learning, offline RL, multi-agent and hierarchical RL, emerging trends and applications.
AIL7023 Graph Machine Learning | 4 credits (3-0-2)
Introduction and Motivations for Machine Learning with Graphs, Graph and its Matrices, Graph Theory, Laplacian Matrix and Smoothness, Graph Learning from Data, Graph Embedding, Label Propagation, Manifold Regularization, semi, (un,) and Supervised Learning with Graphs, Graph Data Mining, Deep Learning with Graphs.
AIL7024 Machine Learning | 4 credits (3-0-2)
Nature of intelligence and learning; statistical foundations and the frequentist-Bayesian divide; notions of tasks, models, and features; linear regression and classification models; kernels and SVMs; decision trees and random forests; naive Bayes, logistic regression, and neural networks; key advances in modern deep learning; k-means, Gaussian mixture models, EM algorithm; PCA and autoencoders; basics of statistical learning theory.
AIL7025 Introduction to Artificial Intelligence | 4 credits (3-0-2)
Introduction, problem solving with search, game trees, constraint satisfaction, logic and knowledge representation, planning, bayesian networks, Markov decision processes (MDPs), reinforcement learning, partially observable MDPs, statistical learning, neural networks, generative AI, emerging trends/applications, ethical and societal impact.
AIL7026 Data Mining | 4 credits (3-0-2)
Association rule mining (Apriori algorithm, FP-growth), frequent subgraph mining (gSpan and FSG), clustering (agglomerative clustering, k-means, k-medoid, DBSCAN, OPTICS, MCL, spectral clustering), data streams (FM sketch, reservoir sampling, bloom filters, top-k counting), anomaly detection (chi-square tests, p-values, T-tests), search (k-NN queries, range queries, metric distance, curse of dimensionality, KD-tree, range tree, R-tree, best-first search, minhash, locality sensitive hashing), web mining (PageRank, random walks with restarts, ), graph neural networks and embeddings(Lipschitz embeddings, GCN, GraphSage, GAT, GIN).
AIL7031 Deep Learning for Computer Vision | 4 credits (3-0-2)
Image formation and representation, edge detection, feature detectors, review of deep learning, image classification, semantic segmentation, object detection, video understanding, self-supervised computer vision, domain adaptation, generative learning in computer vision.
AIL7032 Natural Language Processing | 4 credits (3-0-2)
NLP concepts: tokenization, syntax, information extraction, question answering, semantics, text classification, reasoning and textual inference, language modeling, discourse, machine translation, instruction following, knowledge-based NLP, NLP for low-resource languages, Text2Code.CS techniques: Word2Vec, LSTMs, key-value attention, Transformers, LLMs, pre-training, fine-tuning, LoRA, instruction fine-tuning, RLHF, LRMs, RLVR, Efficient LLMs, RAG, agentic AI.
AIL7051 Bayesian Statistical Modelling and Computation | 4 credits (3-0-2)
The course will cover the use of Bayesian statistical models for a range of standard settings including regression, causal inference, and generalised linear models; in this context it will introduce key conceptual ideas like sampling, inference, confounding, Bayesian regularisation and model comparison, and also draw contrasts with frequentist ways of thinking about these notions.
AIL7061 Computational Cognition | 3 credits (3-0-0)
The course will provide a background about the historical ideas and main presuppositions of AI and computational cognitive science. It will discuss the basics of representation, which is the dominant approach to studying both human and machine cognition. Then, it shall delve into details about computational models, functional architectures, cognitive architectures and also discuss more recent advances linking discrete with analog systems.
AIL7062 AI for Computational Biology | 4 credits (3-0-2)
The course will be an attempt to look at how ideas and tools from AI and computational modelling can be applied to understand the workings of living organisms: including specific topics such as computational genomics, phylogenetics, evolutionary modelling, gene regulatory networks and systems biology, biological graphs and networks, and structural biology.
AIL7063 Artificial Intelligence for Earth Observation | 4 credits (3-0-2)
Introduction to earth observation data, challenges for artificial intelligence in earth observation data analysis, different learning paradigms, semantic segmentation of satellite images, change detection, aleatoric and epistemic uncertainty in earth observation data and models, deep learning model explainability, deep learning for time-series data analysis, advanced architectures for hyperspectral images, multi-sensor and multi-modal adaptation and fusion, vision foundation and regression models for earth observation applications. Practical using relevant programming language and deep learning library for multi-spectral and hyperspectral image classification, and multi-sensor analysis.
AIL8001 Mathematics of Machine Learning | 3 credits (3-0-0)
Normed linear space, Banach space, Hilbert space, orthonormal basis, projection theorem, dual spaces, Riesz representation theorem, bounded linear operators, compact operators, spectral theorem for self-adjoint compact operators. Reproducing kernel Hilbert space (RKHS), positive definite functions, feature maps, Gaussian probabilistic inequalities, Tikhonov-type regularization, general regularization, representer theorem, convergence analysis of empirical risk minimization, adaptive regularization parameter choice rules. Approximation by trigonometric polynomials, localized kernel approximations, neural networks, RBF neural networks, universal approximation, convergence analysis of neural networks.
AIL8021 Special Topics in Machine Learning I | 4 credits (3-0-2)
Will cover some advanced and emerging topics in the area of machine learning, which are likely to evolve with time; examples as of now would include transformer architectures and large language models, deep generative models, advanced optimisation techniques, probabilistic graphical models, graph neural networks, model compression and knowledge distillation, etc.
AIL8022 Special Topics in Machine Learning II | 3 credits (3-0-0)
Will cover some advanced and emerging topics in the area of machine learning, which are likely to evolve with time; examples as of now would include transformer architectures and large language models, deep generative models, advanced optimisation techniques, probabilistic graphical models, graph neural networks, model compression and knowledge distillation, advanced computational learning theory, etc.
AIL8023 Special Topics in Reinforcement Learning | 3 credits (3-0-0)
To be decided on a case-by-case basis by the instructor.
AIL8024 Computational Learning Theory | 3 credits (3-0-0)
We will seek to cover the key theoretical ideas which help us to understand the successes (and failures) of machine learning, and also point out the (many) gaps in our current understanding. The theoretical frameworks/notions discussed will include empirical risk minimisation, VC dimension, PAC learning, Rademacher complexity, the minimum description length principle, no-free-lunch theorems, statistical invariants, and some ideas from probability theory, information theory, and algebraic topology. We will also look at how learning theory can have practical utility in terms of error and generalisation bounds on the performance of learning algorithms, which in turn may help us to design better models and also address issues of data sufficiency.
AIL8027 Advanced Reinforcement Learning | 4 credits (3-0-2)
In this course, we will delve into the intricacies of reinforcement learning (RL) by exploring the advanced topics in the field. RL, a sprawling research area, holds promise for applications in diverse real-world domains such as robotics, autonomous driving, smart transportation, finance, supply-chain logistics, training LLMs, games etc. However, the challenges we confront in these domains often do not align with ideal conditions, necessitating a departure from simply applying our preferred off-the-shelf RL algorithms. For instance, we may encounter scenarios with multiple learning agents in the environment, sparse reward structures, multiple dynamic goals, incorporating constraints in the policy optimization etc. We will also cover recent crucial applications of RL in training LLMs.
Topics: Multi-agent RL, constrained RL, hierarchical RL, RL for LLMs, imitation learning, unsupervised RL, distributional RL, meta RL, goal conditioned RL and other advanced topics in RL.
Topics: Multi-agent RL, constrained RL, hierarchical RL, RL for LLMs, imitation learning, unsupervised RL, distributional RL, meta RL, goal conditioned RL and other advanced topics in RL.
AIL8032 Large Language Models | 3 credits (3-0-0)
Course introduction, introduction to natural language processing, introduction to language models: statistical language modeling, introduction to language models: advanced smoothing and evaluation, word vectors, neural language models, introduction to transformer (self-attention, multi-head attention positional encoding, layer normalization), pre-training strategies, advanced attention, Tokenization strategies, mixture of experts, scaling laws, pre-training of causal LMs and in-context learning, instruction tuning, alignment of language models (PPO, DPO, GRPO, etc.), parameter efficient fine-tuning, quantization, pruning and distillation, efficient LLM decoding, retrieval-based language models, multimodal models, tool augmentation, reasoning in LLMs, long context LLMs, knowledge editing, self-evolving LLMs, an alternate formulation of transformers, bias and interpretability.
AIL8041 Special Topics in Data Science | 3 credits (3-0-0)
Recent advances in big-data management, applications of machine learning to understand structured and unstructured data in large enterprises, human-centric analysis of large-data, spatio-temporal analytics, data cleaning and data integration using machine learning etc.
AIL8061 Special Topics in AI Applications | 3 credits (3-0-0)
The course will focus on one or two of the specialized applications of AI such as computational biology, genetics, healthcare, wellness, transportation systems, software development, hardware design, recommendation systems, music, etc. The course will discuss the various AI techniques developed for problems in these domains.
AIL8062 Special Topics in Computer Vision | 3 credits (3-0-0)
The course will focus on few specialized areas of computer vision such as earth observation, mobility, transportation systems, healthcare, wellness, virtual and augmented reality, computational biology, genetics, etc. The course will discuss various computer vision techniques developed for problems in these domains.
AIN7091 Teaching/Research Practicum | 4 credits (0-0-8)
Students are expected to assist with teaching that includes preparing course materials, managing class activities and examination, providing support to student engagement, grading assignments. They are also expected to help with practical, tutorials, class projects (depending on the course content) and other administrative tasks related to the course and the school.
AIV7090 Ethical Considerations in MINDS | 1 credits (1-0-0)
Introduction to ethical issues in AI: bias, fairness, transparency, reliability, accountability, job loss, privacy, human-aware AI. AI models to debias training data. Explainability in ML. Adversarial robustness. Federated learning.
AIV7091 Research Communication | 1 credits (1-0-0)
Introduction to research communication, poster preparation, oral presentation, and paper writing. Literature review - methodologies and standard tools. Research methodology - problem formulation, design/implementation, evaluation.
AIV8001 Special Module in Machine Learning | 1 credits (1-0-0)
Will cover some specialised and emerging theme in machine learning research or practice, which may evolve with time as per the latest trends and interests. Examples as of now may include the interface of information theory and deep learning; the relevance of statistical physics ideas to understanding emergent behaviours in machine learning systems; specialised deep learning architectures for particular tasks or data sets, etc.
AIV8061 Special Module in AI Applications | 1 credits (1-0-0)
The course will focus on some specialized applications of AI such as computational biology, genetics, healthcare, wellness, transportation systems, software development, hardware design, recommendation systems, music, etc. Exact content will vary from offering to offering.