Learning structure when labels are absent: components and latent factors, autoencoders, likelihood-based generative models, autoregressive image/audio models, PixelCNN variants, and normalizing flows. Built from the PDFs, PPTX and solution docs in udl/.
Previous-year EC-2 regular/makeup problems worked end-to-end with collapsible answers and missing-figure method notes.
01 · cheatsheetDense one-glance revision card: PCA/ICA/CCA/LLE, autoencoders, autoregressive models, PixelCNN variants and flows.
02 · slides explainedLecture-by-lecture guide in plain language, plus the first-principles companion notes for the same arc.
03 · question bankQuestions only, no answers, grouped by lecture/topic for active recall before opening solved paper.
04 · formula sheetEvery equation likely to appear: PCA, autoencoder losses, BCE, convolutions, MLE, autoregressive likelihood and flow Jacobians.
05 · book explainedCourse-book map for Understanding Deep Learning Ch16 plus Geron Ch8 and autoencoder/flow reading themes.
06 · referencesTextbooks, papers, blogs and source map used for the UDL vault.
exam guideAdditional guide based on previous-year papers, Lecture 8 tips and slide emphasis: what to revise, what to practice, how to answer.
★ · SOLVED PAPER
Previous-year EC-2 regular and makeup questions, plus Lecture 8 assignment-style drills, worked with collapsible answers.
03 · QUESTION BANK
Questions only, no answers, grouped by topic. Use this for active recall before opening the solved paper.
EXAM GUIDE
What to prioritize from slides and previous-year papers: likely numericals, conceptual comparisons, and answer patterns.