Artificial Intelligence

9

Discrete Mathematics for AI

Builds the sets, relations, functions, combinatorics, recurrence-relation, and Boolean-algebra vocabulary that underlies data structures, search-space sizing, and rule-based reasoning in AI, worked through concrete numeric examples.

› 10

Linear Algebra for AI

Builds vectors, matrices, and tensors as representations and transformations, then walks through eigenvalues, eigenvectors, and matrix decomposition using concrete worked numeric examples relevant to AI systems.

› 11

Calculus for Machine Learning

Builds derivatives, gradients, Jacobians, Hessians, and the multivariable chain rule as the calculus toolkit that powers gradient-based learning inside neural networks and other modern AI systems today.

› 12

Probability Theory for AI

Introduces random variables, probability distributions, conditional probability, expectation, and variance as the mathematical language AI systems use to reason carefully under genuine, irreducible uncertainty.

› 13

Statistics and Statistical Inference

Covers estimation, sampling distributions, confidence intervals, hypothesis testing, and regression as the inferential tools for judging whether a measured AI result is genuinely statistically meaningful.

› 14

Bayesian Methods

Develops Bayes' theorem, priors, likelihoods, posterior updating, and Bayesian decision theory as the framework AI systems use to revise beliefs rationally as new evidence arrives.

› 15

Optimization for Artificial Intelligence

Covers convexity, gradient descent, constrained optimization, stochastic optimization, and duality as the mathematical basis for how AI systems efficiently search for good model parameters at scale.

› 16

Information Theory

Introduces Shannon entropy, cross-entropy, mutual information, and information gain as the shared mathematical toolkit behind data compression and feature selection in machine learning.

› 17

Graph Theory and Network Mathematics

Covers graphs, vertices, edges, paths, connectivity, trees, and traversal algorithms as the shared mathematical language behind routing, knowledge graphs, and graph-structured neural networks.

› 18

Decision Theory and Utility

Develops rational-preference axioms, utility functions, expected utility maximization, and risk attitudes as the normative framework decision-theoretic AI agents use to choose actions under uncertainty.

› 19

Game Theory and Strategic Decision-Making

Introduces normal-form games, dominant strategies, and Nash equilibrium, plus cooperative bargaining and mechanism design, as the framework for reasoning about multi-agent AI systems whose payoffs depend on each other's choices.

› 20

Causal Inference and Counterfactual Reasoning

Introduces causal directed acyclic graphs, confounding, the do-operator for interventions, and structural counterfactuals as the formal tools for reasoning about cause and effect beyond correlation.

› 21

Programming for Artificial Intelligence

Teaches Python-based computational thinking, algorithm implementation, systematic debugging with pdb, and testing with unittest as the programming foundation every later AI topic builds on.

› 22

Data Structures and Algorithms

Covers core data structures — arrays, hash tables, trees, and graphs — plus sorting, searching, algorithmic complexity, and how to argue an algorithm is correct rather than merely lucky.

› 23

Software Engineering for AI

Covers modular software design, API design, documentation conventions, automated testing, and version control as the engineering discipline that keeps AI research code maintainable and trustworthy.

› 24

Databases and Data Management

Covers relational data modeling and SQL, indexing and ACID transactions, and NoSQL and vector database models, building the vocabulary to choose storage architecture deliberately for AI systems.

› 25

Data Engineering for AI

Covers data pipeline design — collection, ingestion, transformation, validation, and storage — plus data provenance and lineage as the discipline that keeps AI training data trustworthy and traceable.

› 26

Parallel, Distributed, and Cloud Computing

Explains how GPUs, clusters, containers, orchestration, and cloud services let AI workloads scale computation across many processors and machines efficiently.

› 27

Data Preparation and Exploratory Analysis

Covers how to clean messy data — missing values, outliers, duplicates — and use visualization and transformation to explore and understand a dataset before modeling.

› 28

Feature Engineering and Representation

Explains how raw data is turned into model-ready numeric features through encoding, scaling, selection, transformation, and dimensionality reduction.

› 29

Dataset Design, Annotation, and Quality

Covers sampling and class balance, annotation and inter-annotator agreement, and documentation frameworks that make datasets trustworthy and auditable.

› 30

Privacy-Preserving Data and Synthetic Data

Covers anonymization, consent, data minimization, differential privacy, and synthetic data generation as approaches to sharing useful data while protecting individuals.

› 31

State-Space Representation and Problem Solving

Introduces the formal state-space model of problem solving — states, actions, transition models, goal tests, and path costs — that every classical search algorithm operates on.

› 32

Uninformed Search Algorithms

Covers the uninformed (blind) search strategies — breadth-first, depth-first, uniform-cost, depth-limited, and iterative-deepening search — that explore a state space using only its structure.

› 33

Heuristic and Informed Search

Introduces informed search — greedy best-first search and A* — along with heuristic design, admissibility, consistency, and memory-bounded variants like IDA*.

› 34

Local and Metaheuristic Optimization

Covers local and metaheuristic optimization — hill climbing and its variants, simulated annealing, tabu search, beam search, and other stochastic optimization strategies.

› 35

Constraint-Satisfaction Problems

Explains how to formulate assignment and scheduling problems as constraint-satisfaction problems and solve them efficiently with backtracking search, ordering heuristics, and constraint propagation.

› 36

Adversarial Search and Game Playing

Covers how agents choose optimal moves against adversaries using minimax and alpha-beta pruning, evaluation functions for practical depth limits, and Monte Carlo tree search for games with huge branching factors.

› 37

Automated Planning and Scheduling

Introduces classical planning representations (STRIPS, PDDL), planning-graph and hierarchical planning techniques, and how a validated plan is scheduled against limited time and resources.

› 38

Planning and Decision-Making Under Uncertainty

Explains how planning extends beyond the classical fully-observable, deterministic setting to contingent plans, probabilistic MDP-based planning, partial observability, and execution-time replanning.

› 77

Generative AI

An introduction to generative modeling, contrasting it with discriminative modeling, and a tour of three foundational architectures: GANs, VAEs, and diffusion models.

› 80

Natural Language Processing

Traces NLP from the classic annotation pipeline of tokenization, tagging, and parsing to distributional word embeddings and today's dominant Transformer-based large language models.

› 84

Computer Vision

An introduction to how machines interpret images, covering the classification, detection, and segmentation task taxonomy plus the ImageNet-driven rise of deep convolutional networks.

› 96

Large Language Models

An introduction to the Transformer architecture and the pretrain-then-adapt paradigm that together underpin modern large language models like GPT-3 and ChatGPT.

›

Responsible AI

A capstone survey of how organizations and governments manage AI risk, covering the NIST AI RMF and the EU AI Act and OECD AI Principles.

› 39

Logic for Artificial Intelligence

Introduces propositional and first-order logic as formal languages for representing knowledge, covering syntax, semantics, entailment, quantifiers, and the boundary between decidable and undecidable inference.

› 40

Automated Reasoning and Theorem Proving

Covers resolution refutation, unification and the most general unifier, and the forward- and backward-chaining strategies that let automated systems search efficiently for proofs over large rule bases.

› 41

Knowledge Representation

Surveys how AI systems formally represent objects, categories, properties, events, actions, time, space, and other agents' beliefs — the ontological building blocks beneath any reasoning system.

› 42

Ontologies and Semantic Technologies

Covers taxonomies, description logics, and the W3C semantic web stack — RDF triples, OWL ontologies, and SPARQL queries — as the standards that let machines share and reason over structured knowledge.

› 43

Knowledge Graphs

Introduces knowledge graphs as directed labeled graphs of entities and relations, covering how they are built and curated, how embeddings turn discrete triples into vectors for learning, and how querying, completion, and reasoning operate over inherently incomplete graphs.

› 44

Expert Systems and Rule-Based AI

Covers the architecture of expert systems — knowledge bases and production rules, forward- and backward-chaining inference engines, explanation facilities, and the knowledge-acquisition process — through the landmark case of MYCIN.

› 45

Commonsense and Non-Monotonic Reasoning

Examines reasoning that draws default conclusions from incomplete information and revises them when exceptions appear, covering circumscription, default logic, belief revision, and temporal reasoning as formal answers to the problem of non-monotonicity.

› 46

Fuzzy Logic and Approximate Reasoning

Introduces fuzzy sets, membership functions, and linguistic variables, then walks through Mamdani-style fuzzy rule inference and the historic 1975 fuzzy steam-engine controller that first demonstrated the approach's practical value.

› 47

Neuro-Symbolic Artificial Intelligence

Examines how neuro-symbolic AI combines neural networks' pattern-learning strengths with symbolic reasoning's compositional guarantees, using AlphaGeometry's verified Olympiad-level results as a concrete case study while honestly framing the field's genuinely open research challenges.

› 48

Probabilistic Reasoning and Inference

Covers conditional probability, Bayes' rule, conditional independence, and Bayesian network structure, then extends to likelihoods and probabilistic decision-making as the mathematical foundation for reasoning under genuine uncertainty.

› 49

Bayesian Networks and Graphical Models

Introduces directed and undirected probabilistic graphical models, showing how conditional independence structure enables exact inference algorithms like variable elimination, belief propagation, and junction trees, plus principled approximate methods when exact computation becomes intractable.

› 50

Sequential Probabilistic Models

Covers Markov chains, hidden Markov models, dynamic Bayesian networks, and state-space models as the core toolkit for reasoning about hidden state that evolves over time and is only ever observed indirectly.

› 51

Monte Carlo and Sampling Methods

Covers importance sampling, Markov chain Monte Carlo (Metropolis-Hastings and Gibbs sampling), particle filtering, and simulation as tools for approximating intractable probabilistic computations with random samples.

› 52

Probabilistic Programming

Explores how probabilistic programming languages express generative models as executable code with sampling and conditioning primitives, letting general-purpose inference engines automate Bayesian computation that would otherwise require bespoke derivations.

› 54

Supervised Learning: Regression

Covers linear regression and the least-squares objective, regularized regression via ridge and lasso, and nonlinear and probabilistic approaches including locally weighted regression and the maximum-likelihood view of least squares.

› 55

Supervised Learning: Classification

Covers logistic regression as a discriminative classifier, k-nearest neighbors as an instance-based method, and naive Bayes as a generative classifier built on a conditional-independence assumption.

› 56

Decision Trees and Ensemble Learning

Covers how decision trees recursively split data using impurity measures, how bagging and random forests combine many trees to reduce variance, and how boosting and stacking combine diverse models to reduce error further.

› 57

Support Vector Machines and Kernel Methods

Covers the maximum-margin principle behind support vector machines, the kernel trick that lets them draw nonlinear decision boundaries without ever computing a high-dimensional feature mapping explicitly, and support vector regression.

› 58

Unsupervised Learning and Clustering

Surveys the major families of clustering algorithms — partitional k-means, hierarchical, probabilistic mixture models fit with EM, density-based DBSCAN, and spectral clustering — and the differing assumptions each makes about cluster shape and structure.

› 59

Dimensionality Reduction and Latent-Variable Models

Covers linear dimensionality reduction through PCA and factor analysis, probabilistic latent-variable generative models, and nonlinear manifold-learning methods including Isomap, LLE, and t-SNE for compressing and interpreting high-dimensional data.

› 60

Semi-Supervised, Self-Supervised, and Weakly Supervised Learning

Examines three strategies for learning when clean, complete labels are scarce: semi-supervised learning that leverages unlabeled data alongside a small labeled set, self-supervised learning that manufactures its own labels from raw data structure, and weak or programmatic supervision from noisy heuristic sources.

› 61

Transfer, Multi-Task, and Domain-Adaptive Learning

Explains how models reuse learned representations across tasks and domains through transfer learning's formal taxonomy, multi-task learning with shared representations, and domain adaptation techniques for handling distribution shift.

› 62

Few-Shot, Zero-Shot, Meta, and Continual Learning

Surveys techniques for learning from very few or zero labeled examples, gradient-based meta-learning algorithms like MAML, and strategies for retaining prior knowledge while learning new tasks without catastrophic forgetting.

› 63

Online, Streaming, and Active Learning

Covers algorithms that learn from data arriving one example at a time, detect and adapt to distributional change (concept drift), and strategically choose which examples are most worth labeling.

› 64

Time-Series Learning and Forecasting

Introduces how to decompose and model temporal patterns in sequential data, classical forecasting methods including ARIMA and exponential smoothing, and how to rigorously evaluate forecast accuracy.

› 65

Anomaly Detection and Imbalanced Learning

Covers detecting rare and novel events, the taxonomy of anomaly types, techniques for learning from severely imbalanced classes, and how to fold asymmetric misclassification costs into model decisions.

› 66

Recommender and Ranking Systems

Covers collaborative filtering and content-based recommendation, matrix-factorization and learning-to-rank methods for ordering items, and the feedback loops and biases that recommender systems create once their own output shapes future training data.

› 67

Machine-Learning Theory

Covers PAC learning and sample complexity bounds, VC dimension as a measure of hypothesis-class capacity, and the bias–variance trade-off that governs generalization from finite training data.

› 68

Model Evaluation and Experimental Design

Covers evaluation metrics, cross-validation and baselines, probability calibration, statistical significance testing for comparing learning algorithms, and systematic error analysis.

› 69

Robust and Causal Machine Learning

Covers distribution shift and out-of-distribution detection, causal discovery methods for uncovering structure from observational data, and treatment-effect estimation via propensity scores and related adjustment techniques.

› 71

Deep-Network Training and Optimization

How to actually get a deep network to train well: weight initialization and batch normalization, regularization via weight decay and dropout, and diagnosing training runs with loss curves, update ratios, and adaptive optimizers like Adam.

› 72

Convolutional Neural Networks

A deeper look at convolutional networks: convolution arithmetic and receptive fields, LeNet-5's pioneering architecture, how feature hierarchies and spatial structure emerge from stacked layers, and how residual connections solved the degradation problem in very deep networks.

› 73

Recurrent and Sequence Neural Networks

Introduces recurrent neural networks, the vanishing and exploding gradient problem in backpropagation through time, and how LSTM and GRU gating mechanisms let networks retain and forget information across long sequences.

› 74

Attention Mechanisms and Transformers

Covers the attention mechanism's alignment weights, the Transformer's self-attention and multi-head attention, positional encoding, and the encoder-decoder architecture that replaced recurrence in modern sequence models.

› 75

Representation, Embedding, and Contrastive Learning

Explores distributed representations and embeddings, metric and contrastive learning objectives, and how self-supervised contrastive frameworks like SimCLR learn transferable representations without labels.

› 76

Autoencoders and Variational Autoencoders

Explains autoencoder architectures for unsupervised compression and reconstruction, and how variational autoencoders add a probabilistic latent space enabling smooth interpolation and controlled generation.

› 78

Graph Neural Networks and Geometric Deep Learning

Covers why graph-structured data breaks the grid and sequence assumptions behind CNNs and RNNs, how message-passing and graph convolutional networks learn from graphs instead, and why stacking too many GNN layers causes node representations to over-smooth.

› 79

Efficient and Robust Deep Learning

Covers how pruning, quantization, and knowledge distillation shrink deep networks for efficient deployment, and how the discovery of adversarial examples led to robust-optimization defenses in an ongoing, still-unresolved arms race.

› 81

Language Modelling, Translation, and Summarization

Traces language modeling from n-gram statistics through recurrent and attention-based neural architectures, showing how the same core ideas power machine translation, abstractive summarization, and open-ended text generation.

› 82

Information Retrieval, Question Answering, and Dialogue

Covers the probabilistic foundations of search and ranking, extractive and retrieval-augmented approaches to question answering, and the evolution of dialogue systems from rule-based pattern matching to retrieval-grounded conversational AI.

› 83

Speech and Audio Intelligence

Surveys the acoustic modeling pipeline behind automatic speech recognition, neural approaches to speech and audio synthesis, and the broader landscape of speaker processing, audio classification, and music AI.

› 85

Object Detection, Segmentation, and Tracking

Covers the evolution of object detection from region-proposal to single-pass architectures, pixel-level segmentation with encoder-decoder networks, and the algorithms that link per-frame detections into consistent object tracks over video.

› 86

3D Vision, Scene Understanding, and Video Intelligence

Examines how AI systems recover 3D structure and camera motion from images, estimate human pose within a scene, and interpret actions and events unfolding across video.

› 87

Multimodal Artificial Intelligence

Explores how AI systems align and jointly reason across text, images, audio, video, and other modalities, from early multimodal taxonomies and benchmarks through large-scale contrastive vision-language pretraining and its extensions.

› 88

Reinforcement Learning

An advanced introduction to the agent-environment MDP framework, the reward hypothesis, and Deep Q-Networks, the deep RL method that learned to play Atari games from raw pixels.

› 89

Value-Based and Policy-Based Reinforcement Learning

Covers value-based control via temporal-difference learning and Q-learning, direct policy optimization via the policy gradient theorem and REINFORCE, and actor-critic methods that combine both, tracing the path to the deep RL systems — DQN, TRPO/PPO, AlphaGo — that scaled these ideas to large, high-dimensional problems.

› 90

Advanced Reinforcement Learning

Surveys reinforcement learning beyond the standard single-agent online MDP: model-based planning and offline learning from a fixed dataset, hierarchical decomposition of long-horizon tasks, imitation and inverse reinforcement learning from expert demonstrations, multi-agent learning in shared environments, and the open safety problems that arise once RL systems act in the real world.

› 91

AI Agents

Traces the intelligent-agent concept from Russell and Norvig's classical reflex, model-based, goal-based, and utility-based taxonomy to modern LLM agents that interleave reasoning and tool use.

› 92

Multi-Agent Systems and Collective Intelligence

Extends single-agent taxonomies into multi-agent settings: cooperative, competitive, and mixed-motive interaction, Nash equilibrium and Axelrod's iterated Prisoner's Dilemma tournaments as a model of decentralized cooperation, and multi-agent reinforcement learning (independent learners, MADDPG, COMA, AlphaStar, emergent competition and communication).

› 93

Robotics

An introduction to how robots sense, plan, and act, how kinematics and dynamics govern manipulator control, and how simulation-trained reinforcement learning transfers to physical hardware.

› 94

Robot Perception, Localization, Mapping, and Planning

Covers how mobile robots estimate their own state from noisy sensors using Bayes, Kalman, and particle filters; how they build maps of unknown environments while localizing within them via SLAM and occupancy grids; and how they plan collision-free paths using A* search and sampling-based motion planners like RRT.

› 95

Robot Learning and Human–Robot Interaction

Covers how robots learn manipulation skills from demonstration and joint perception-control training, how sim-to-real transfer techniques like automatic domain randomization and dynamics randomization generalize beyond a single hand or task, and how human-robot interaction research defines the dimensions and safety strategies of effective, safe collaboration between people and robots sharing physical space.

› 97

Generative AI Adaptation, Alignment, and Evaluation

Covers how pretrained language models are adapted into instruction-following assistants through instruction tuning and human-preference learning — from RLHF and Direct Preference Optimization to Constitutional AI — and how the field tries, imperfectly, to measure factuality and overall quality once that adaptation is done.

› 98

Retrieval-Augmented Generation and Generative AI Applications

Covers how generative systems are grounded in external information and connected to external actions: dense embeddings and vector retrieval, retrieval-augmented generation, tool calling and reasoning-acting loops, and structured and multimodal generation.

› 99

AI Ethics

Examines algorithmic bias through the COMPAS recidivism case study and surveys global AI governance through UNESCO's 2021 Recommendation on the Ethics of Artificial Intelligence.

› 100

AI Engineering, MLOps, Applications, and Capstone Research

The capstone module of the Artificial Intelligence major: how research systems — spanning classical AI, machine learning, language and vision models, agents, and foundation models — become deployed, monitored, and maintained production systems, covering MLOps and technical debt, scalable training and AI hardware economics, and reproducibility; closing with guidance for scoping an independent capstone project that draws on the whole major.

›