Algorithms (2IL15) Lecture 13 Wrap-up lecture 1 TU/e Algorithms - PowerPoint PPT Presentation

TU/e Algorithms (2IL15) – Lecture 13 Algorithms (2IL15) – Lecture 13 Wrap-up lecture 1

TU/e Algorithms (2IL15) – Lecture 13 Part I of the course: Optimization Problems we want to find valid solution minimizing cost (or maximizing profit)  always start by investigating structure of optimal solution backtracking: just try all solutions (pruning when possible) − extra exercises: backtracking algorithms for NP-hard problems Vertex Cover, 3-SAT, Hamiltonian Cycle, … greedy algorithms: make choice that seems best, recursively solve remaining subproblem − proof is crucial; make sure you know proof structure by heart − extra exercises: greedy algorithms explained in book dynamic programming: give recursive formula, fill in table − make sure you know steps to design dynamic-programming algorithm − extra exercises: dynamic-programming algorithms explained in book 2

TU/e Algorithms (2IL15) – Lecture 13 Part II of the course: optimization problems on graphs − 2 single-source shortest paths 1 Find shortest path from 2 1 source vertex to all other vertices 2 all-pairs shortest paths 2.3 4 Find shortest paths between all pairs of vertices maximum flow 4 3 Find maximum flow from 1 5 3 2 source vertex to target vertex 8 3 maximum bipartite matching Find maximum number of disjoint pairs of vertices with edge between them 3

TU/e Algorithms (2IL15) – Lecture 13 single-source / all-pairs shortest paths − you should know and understand the various algorithms SSSP: Bellman-Ford, Dijkstra, algorithm on DAGs APSP: relation to matrix multiplication, Johnson’s reweighting − extra exercises: give algorithms, give proofs of important lemmas maximum flow + maximum bipartite matching − definition of flow network, flow − Ford-Fulkerson method + Edmonds-Karp implementation (concept of augmenting flow, residual path, etcetera) − relation between flows and cuts (e.g. max flow = min cut) − integer capacities: there is always a max flow that is integral  basis for many applications, e.g. for max bipartite matching) 4

TU/e Algorithms (2IL15) – Lecture 13 Part III of the course: miscellaneous topics Part III of the course: Miscellaneous topics NP-completeness − definitions and concepts: P, NP, reductions − list of NP-complete problems you should be able to reproduce NP-completeness proofs … (they also teach you tricks for other NP-completeness proofs) approximation algorithms − definitions and concepts: approximation factor, (F)PTAS − how to prove approximation ratio: compare to lower (upper) bound linear programming − definitions and concepts: LP, standard form integer LP, 0/1-LP − how to model a problem as an LP 5

TU/e Algorithms (2IL15) – Lecture 13 The exam  most exercises are of the same type as the homework exercises  about 1/3 of exercises is (almost) identical to homework exercises  exam may also have some questions where the answer can be found literally on the slides or in the book (reproduce algorithm or proof, definitions) How to prepare 1. study the slides (only look into the book when there are things in the slides that you do not understand) 2. solve all the homework exercises again ! 3. reproduce algorithms / proofs from the book 6

TU/e Algorithms (2IL15) – Lecture 13 What we did not do in the course … … lots, in particular … algorithms for big data geometric algorithms algorithms for geographic data 7

TU/e Algorithms (2IL15) – Lecture 13 Algorithms for Big Data ALU registers L1 cache L2 cache L3 cache internal memory disk Data, data, everywhere. Economist 2010 8

TU/e Algorithms (2IL15) – Lecture 13 I/O- and caching behavior analysis of running time: (asymptotically) determine number of elementary operations as a function of input size basic assumption: elementary operations take roughly same amount of time is this justified? ALU Not when data is stored on disk: registers disk operations can be factor 100,000 (or more) L1 cache slower than operations on data in main memory L2 cache L3 cache − how can we design I/O-efficient algorithms? internal memory − how should we analyze such algorithms? disk − can / should we also worry about caching ? 9

TU/e Algorithms (2IL15) – Lecture 13 streaming algorithms data stream model:  data is read sequentially in one (or few) passes  size of the working memory is much smaller than size of data example: traffic going through high speed routers ( data can not be revisited ) − how can we design algorithms for data streams? (e.g. maintain approximate median of stream of numbers) − how should we analyze such algorithms?  take Advanced Algorithms (big data + more approximation algorithms) 10

TU/e Algorithms (2IL15) – Lecture 13 Geometric algorithms 11

TU/e Algorithms (2IL15) – Lecture 13 computer-aided design and manufacturing (CAD/CAM)  motion planning  virtual walkthroughs 3D model of power plant (12.748.510 triangles) 12

TU/e Algorithms (2IL15) – Lecture 13 3D copier 9 surface reconstruction 5 3 4 6 7 8 2 1 3D printer 3D scanner 13

TU/e Algorithms (2IL15) – Lecture 13 many areas where geometry plays important role:  geographic information systems  computer-aided design and manufacturing  robotics and motion planning  computer graphics and virtual reality  databases  structural computational biology  and more … salary 50.000 30.000 30 45 age 14

TU/e Algorithms (2IL15) – Lecture 13 Geometric data structures  store geometric data in 2-, 3- or higher dimensional space such that certain queries can be answered efficiently  applications: GIS, graphics and virtual reality, … Examples:  point location  range searching  nearest-neighbor searching want to know more about algorithms and data structures for spatial data?  take Geometric Algorithms 15

TU/e Algorithms (2IL15) – Lecture 13 Algorithms for Geographic Data

TU/e Algorithms (2IL15) – Lecture 13 geographic information systems (GIS) Mississippi delta Arlington Arlington Memphis Memphis Memphis Memphis 30 cities: 4 30 = 10 18 possibilities 17

TU/e Algorithms (2IL15) – Lecture 13 automated cartography examples:  map labeling  scaling and simplification  schematization  maps for various types of data, e.g., flows 18

TU/e Algorithms (2IL15) – Lecture 13 trajectory data data of  animals, humans, vehicles, … Examples of analysis tasks:  How similar are two trajectories?  How to find clusters?  How to represent a cluster?  How to simplify a trajectory?  …  take Algorithms for Geographic Data 19

TU/e Algorithms (2IL15) – Lecture 13 THE END well, except for the exam … comments / suggestions for improvements are welcome 20