Randomized and Approximation Algorithms

Advanced Course, 2+2

Basic Information

Given by:Antonios Antoniadis and Marvin Künnemann
Time:Tuesday, 14:00- 16:00 
Exercises by:André Nusser
ExercisesThursday, 14:00-16:00
Room:024 E1.4
First Meeting:October 16th. 
Credits:6 credit points
Grade formula:The grade is fully determined by the exam (depending on number of participants, either oral or written). To qualify for the exam, 50 % points in the exercise sheets are required.
Prerequisites:You will need to be mathematically mature. You should be able to read and understand technical/mathematical texts, and should have basic knowledge in Algorithms and Probability Theory.

News

Description

Several practically relevant algorithmic problems are unfortunately not known to have deterministic efficient algorithms. More specifically, for several important problems, it is highly unlikely that an efficient algorithm exists that produces an optimal solution on every input instance. Since often such problems are too important to be left unadressed, there are several "relaxations" being used to adress such problems. Two of them are:

  • Approximation Algorithms: One can relax the objective of searching for the optimal solution and instead design an efficient algorithm that produces solutions which are provably "close" in value to the optimal one.
  • Randomized Algorithms, and Probabilistic Analysis of Algorithms: Often, allowing an algorithm to make random choices during its execution leads to significantly more efficient computation (possibly with the drawback that the efficiency is only guaranteed with some probability, or that the output is correct only with some probability). Especially in the context of approximation algorithms, such randomized methods provide powerful tools.  Furthermore, many problems are known to be difficult only in specific, pathetic instances whereas for instances apearing in practice efficient algorithms may exist. Probabilistic analysis of algorithms can, in many cases, give a theoretical explanation of this phenomenon.

In this course we will focus on several techniques for designing and analyzing randomized and approximation algorithms. We will also see a couple of interesting recent results in the area.

 

Schedule

DateSpeakerTopicReferenceExercise Sheet
Oct 16

Antonios & Marvin

Introduction to the Course. Introduction to Approximation Algorithms. Greedy Algorithms

W&S: Subsect. 1.1, 1.2 (first part), 1.6 (first part), 2.1, 2.2Exercise Sheet 1
Oct 23MarvinIntroduction to Randomized Algorithms. Max-Cut, Max-3SAT.  
Oct 30 AntoniosLocal Search  
Nov 6AntoniosDynamic Programming. Knapsack, Bin Packing. Polynomial Time Approximation Schemes.  
Nov 13MarvinConcentration I: Chebychev and Applications  
Nov 20AntoniosLinear Programming, Deterministic Rounding.   
Nov 27MarvinConcentration II: Chernoff Bounds.  
Dec 4MarvinLinear Programming, Randomized Rounding.  
Dec 11AntoniosLinear Programming, Primal-Dual Algorithms.  
Dec 18AntoniosTBD  
Jan 8TBDTBD  
Jan 15MarvinTBD  
Jan 22MarvinTBD  
Jan 29AntoniosTBD  
Feb 5MarvinTBD  

Bibliography

  • D.P. Williamson, D.B. Shmoys. The Design of Approximation Algorithms. Cambridge University Press, 2011.
  • M. Mitzenmacher, E. Upfal. Probability and Computing. Cambridge University Press, 2005.