CS 7870: Seminar in Theoretical Computer Science — Fall 2023

Syllabus      Tentative Course Schedule

Each lecture below will be self-contained. We will prepare notes for every lecture. The associated readings for each lecture are suggested references. These readings will help gain a better understanding of the material covered, and will provide avenues for other explorations.


Date Topic, Readings, and Deliverables
Fri 9/8
Introduction & Clustering
Overview of the course; center-based clustering; algorithms for k-center
Lecture Notes
References: Williamson-Shmoys Section 2.2, Chekuri, Section 9.1;
Tue 9/12
Nets for metric spaces and nearest neighbor search
Lecture Notes (same as above)
Also, introduction to the k-median problem
PS1 out
Fri 9/15
Local Search for k-Median
The k-median problem; bicriteria approximation using LP rounding; analysis of local search
Lecture Notes
Readings: Williamson-Shmoys Section 9.2, Chekuri, Sections 9.3-9.4;
PS1 out
Tue 9/19
Clustering and Perturbation Stability
Perturbation stability; Optimal k-median solution under perturbation stability
Lecture Notes
References: Roughgarden notes; [AMM17]
Fri 9/22
Linear Programming: Introduction and Applications
Formal definitions; formalizing optimization problems and relaxations as LPs; applications; rounding LP relaxations
Lecture Notes
References: Goemans notes (Sections 1-7); Roughgarden notes
Tue 9/26
Linear Programming: Rounding and Geometry
Rounding LP relaxations; Geometry of LP polytopes; extreme point solutions; iterative rounding
Lecture Notes
References: Goemans notes (Sections 7-8)
PS1 due
Fri 9/29
Linear Programming: Duality
Weak and strong duality; Dual-fitting and Primal-Dual Schema
Lecture Notes
References: Goemans notes (Sections 7-8); Williamson-Shmoys (Sections 7.1,7.3)
PS2 out
Tue 10/3
Linear Programming Wrapup
Rounding algorithms and Primal-Dual Schema
Lecture Notes
References: Goemans notes (Sections 7-8); Williamson-Shmoys (Sections 7.1,7.3)
Fri 10/6
Principal Component Analysis
Dimensionality reduction; goal of PCA; use cases; characterizing principal components
Lecture Notes
References: Roughgarden-Valiant notes (one and two)
Tue 10/10
Singular Value Decomposition and Applications
SVD introduction; SVD vs PCA; low-rank approximation; application to clustering; intro to high-dimensional geometry
Lecture Notes
References: Blum-Hopcroft-Kannan (Chapter 3; sections 2.3-2.4, 2.6)
Fri 10/13
Clustering in High-Dimensional Spaces
Geometry of high-dimensional spaces; Unit ball and spherical Gaussians; Clustering mixture of Gaussians
Lecture Notes
References: Blum-Hopcroft-Kannan (Chapter 3; section 2.8)
Tue 10/17
Random Projections
The Random Projection Theorem; Johnson-Lindenstrauss Lemma; Intro to compressed sensing
Lecture Notes
References: Blum-Hopcroft-Kannan (Section 2.7); Fresken (Sections 1.1-1.3) for summary, variants, uses
Fri 10/20
Compressed Sensing
Warmup: discrete sparse recovery; Compressed sensing; Applications
Lecture Notes
References: Musco notes; Roughgarden-Valiant notes
PS2 due
Tue 10/24
Compressed Sensing Wrapup and Intro to Smoothed Analysis
Complete compressed sensing proof; Notions of smoothed analysis and complexity
Lecture Notes
References: Roughgarden notes
Fri 10/27
Smoothed Analysis of Local Search
Analysis of 2-OPT heuristic for TSP
Lecture Notes
References: Roughgarden notes
PS3 out
Tue 10/31
Smoothed Complexity and Pseudo-polynomial Time Algorithms
Equivalence between poly-time smoothed complexity and pseudo-polynomial time algorithms for binary optimization
Lecture Notes
References: Beier-Vocking 2004 PS3 out
Fri 11/3
Online Learning
The online learning framework; regret; analysis of weighted majority
References: Elad Hazan (Chapter 1)
Lecture Notes
Tue 11/7
Rajmohan away; no class
Fri 11/10
Online Learning, continued; basics of convex optimization
The Hedge algorithm; Introduction to convex optimization; gradient descent; step sizes and convergence bounds
Lecture Notes
References: Elad Hazan (Chapter 2)
Tue 11/14
Online Convex Optimization: Stochastic Gradient Descent
Complete analysis of gradient descent
The optimization powering all of deep learning; convergence analysis
Lecture Notes
References: Elad Hazan (Chapter 3)
Fri 11/17
Rajmohan away; no class
Tue 11/21
Project Proposal Presentations
PS3 due
Fri 11/24
Thanksgiving (no class)
Tue 11/28
Online Algorithms and Competitive Analysis
Wrap-up stochastic gradient descent; intro to competitive analysis: ski rental, paging
Readings: Buchbinder-Naor monograph, chapter 3;
Lecture Notes
Nagarajan notes
Fri 12/1
Online Randomized Algorithms
Randomized upper and lower bounds
Lecture Notes
Readings: Buchbinder-Naor monograph, chapter 4
Tue 12/5
Online Primal-Dual Paradigm
Online primal-dual algorithm for set cover; pointers to many applications
Lecture Notes
Fri 12/8
Project Presentations
Tue 12/12
Project Presentations
Fri 12/15
Project Presentations