DSE - Semester 2 – Knowledge Base

The sole mandatory foundation of Semester 2. ISM establishes rigorous statistical reasoning: Descriptive & Inferential Statistics, Discrete and Continuous Probability Distributions (Binomial, Poisson, Normal, Student’s t, Chi-square, F), Central Limit Theorem, Parametric and Non-Parametric Hypothesis Testing, Analysis of Variance (ANOVA), Linear Regression, Time Series Analysis, and Gaussian Mixture Models (GMM) with Expectation-Maximization.

Focuses on foundational machine learning theory and mathematical mechanics: Linear Regression (OLS, Ridge, Lasso), Logistic Regression, Decision Trees (Entropy, Information Gain, Gini index, Pruning), Support Vector Machines (Lagrangian dual, Slack variables, Kernel trick), Ensemble Models (Bagging, AdaBoost, Gradient Boosting, XGBoost), and Unsupervised Learning (K-Means, Hierarchical Clustering, PCA, t-SNE). Excludes neural networks, computer vision, and cloud MLOps.

An alternative to the theoretical ML course that focuses directly on end-to-end applied predictive pipelines. Covers automated feature preprocessing, linear models, SVMs, tree ensembles, an applied introduction to Artificial Neural Networks (MLP, backpropagation), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs/LSTMs), model interpretability (SHAP, LIME), and AI fairness in enterprise deployment.

One of the most practical and industry-relevant electives in the program. Teaches the principles of human visual perception and cognition (Gestalt laws, pre-attentive attributes), effective visual encodings, chart design and selection taxonomy, interactive dashboard design principles, storytelling with data, exploratory vs. explanatory visualization, geospatial visualization, and visualization ethics.

Examines classical, symbolic, and computational AI algorithms: Intelligent Agent architectures, Uninformed Search (BFS, DFS, Uniform Cost Search), Informed Heuristic Search (A*, IDA*), Adversarial Game Playing (Minimax, Alpha-Beta pruning), Constraint Satisfaction Problems (CSPs, Backtracking, Forward Checking, AC-3), Knowledge Representation using Propositional Logic and First-Order Logic (FOL, Unification, Resolution Refutation), Probabilistic Reasoning (Bayesian Networks, Conditional Independence), Reinforcement Learning (MDPs, Q-Learning), and Fuzzy Logic systems. Excludes deep learning and computer vision.

Dedicated to the data infrastructure that powers production machine learning systems. Covers the Modern Data Stack, scalable data pipelines (Batch ETL vs. Streaming ELT), data versioning architectures (DVC), Feature Stores (Feast architecture), storage formats (Apache Parquet, Apache Arrow, Delta Lake), distributed data processing (Apache Spark for ML), data lineage, metadata management, and AI governance & privacy compliance.