2015 Data Compression Conference Salt Lake City United
- Slides: 34
2015 Data Compression Conference Salt Lake City, United States, 7 -9 April 2015 FASTER COMPRESSED QUADTREES Travis Gagie Javier González Susana Ladra Gonzalo Navarro Diego Seco
2 / 34 Outline 1. Introduction 2. Compressed quadtree representation 3. Experimental evaluation 4. Conclusions 5. Future work
3 / 34 Outline 1. Introduction 2. Compressed quadtree representation 3. Experimental evaluation 4. Conclusions 5. Future work
4 / 34 Introduction • Storing and querying two-dimensional points sets is important • Computational geometry • Geographic information systems • Graphics • Many other fields
5 / 34 Introduction • Real world point sets tend to be clustered • Using a machine word for each point is wasteful • Compressing data allows for faster operations because of caching • We are looking for a structure able to take advantage of the clustering property to improve space and time performance
6 / 34 Introduction • Example with k = 2 0 1 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T = 101111010100100011001000000101011110 L = 01000011001010101000011000100100
7 / 34 Introduction • Retrieving neighbors lists: • Top-down traversal of the tree • Example: retrieve the direct neighbors of page 10 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 0 1 0 0 0 0 1 0 0 1 1 1 0 0 0 2 0 0 0 0 3 0 0 0 0 4 0 0 0 0 5 0 0 0 0 6 0 0 0 0 7 0 0 0 1 0 0 0 0 0 8 0 0 0 1 0 0 0 0 9 0 0 0 1 0 1 0 0 0 10 0 0 0 11 0 0 0 0 12 0 0 0 0 13 0 0 0 0 14 0 0 0 0 15 0 0 0 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30… T = 101111010100100011001000000101011110 children(9) = rank children(2) = rank 1(T, 9)*k (T, 2)* k 22 = 7*4=28 = 2*4=8 L = 01000011001010101000011000100100
8 / 34 Outline 1. Introduction 2. Compressed quadtree representation 3. Experimental evaluation 4. Conclusions 5. Future work
9 / 34 Compressed quadtree representation • Suppose we have a 4 x 4 matrix where each cell contains 0 or 1 • Only 1’s are relevant • Storing only cells with value 1 is necessary • (0, 0) • (3, 0) • (1, 2) • (2, 2) • (1, 3) How can we store them to take advantage of the clustering property?
10 / 34 Compressed quadtree representation • Quadcode: Binary representation of a point in a quadtree Path to a leaf Matrix Quadtree • Closed points share a common prefix (related with quadrants) • We will use this property for compression
11 / 34 Compressed quadtree representation Same than using binary representation and interleaving bits: 1001 1011 We can compute each quadcode in O(1) with precomputed tables
12 / 34 Compressed quadtree representation • Storing points in a trie
13 / 34 Compressed quadtree representation •
14 / 34 Compressed quadtree representation • Heavy path decomposition • Technique for decomposing a rooted tree into a set of heavy paths • Each non-leaf node selects one heavy edge • A heavy edge is the edge of a subtree with the most number of leaves • Selected edges form a heavy path
15 / 34 Compressed quadtree representation • In each step, one edge is selected and the other one is stored in a queue • After finish building a heavy path, a node is retrieved from the queue and the process start again
16 / 34 Compressed quadtree representation
17 / 34 Compressed quadtree representation • Each Next bitmap supports rank operations
18 / 34 Compressed quadtree representation •
19 / 34 Compressed quadtree representation • Remaining quadcode: 100 Remaining path: 0100000
20 / 34 Compressed quadtree representation • Remaining quadcode: 00 Remaining path: 00
21 / 34 Compressed quadtree representation •
22 / 34 Outline 1. Introduction 2. Compressed quadtree representation 3. Experimental evaluation 4. Conclusions 5. Future work
23 / 34 Experimental evaluation • We use grid datasets from different domains: • Geographic information systems (GIS) • Social networks (SN) • Web graphs (WEB) • RDF (RDF) • Space is measured in bits per point (bpp)
24 / 34 Experimental evaluation •
25 / 34 Experimental evaluation • Types of membership • Empty cells • Filled cells • Isolated filled cells • Measure: Average time per query in nanoseconds
26 / 34 Experimental evaluation
27 / 34 Experimental evaluation Left point – Dense Middle point – Medium Right point – Sparse
28 / 34 Experimental evaluation Left point - dblp 2011 Right point – enwiki 2013
29 / 34 Experimental evaluation
30 / 34 Outline 1. Introduction 2. Compressed quadtree representation 3. Experimental evaluation 4. Conclusions 5. Future work
31 / 34 Conclusions • Nice theorical bounds and it is practical • Space requirements are similar to other space- efficient representation of quadtrees • Faster handling isolated filled cells
32 / 34 Outline 1. Introduction 2. Compressed quadtree representation 3. Experimental evaluation 4. Conclusions 5. Future work
33 / 34 Future work • Generalization to higher dimensions • Experimental results for range reporting (querying for points in a particular region)
2015 Data Compression Conference Salt Lake City, United States, 7 -9 April 2015 FASTER COMPRESSED QUADTREES Travis Gagie Javier González Susana Landra Gonzalo Navarro Diego Seco If you have any question: javigonzalez@udec. cl
- Salt lake city compression
- Tracy swena
- Optumlink
- Salt lake city residential building codes
- Salt lake city fsdo
- Va fiduciary hub
- Optum salt lake city
- Fatpipe vpn
- 866-556-8166
- Salt lake temple baptismal font
- Great salt lake salinity
- Osha salt lake technical center
- Data compression in data mining
- Littoral limnetic
- Lake powell and lake mead
- Lake powell and lake mead
- Bread basket of the us
- Led conference 2015
- Ulvan
- Insight 2015 conference
- Clear lake water authority
- Confidential do not distribute
- What is lossy data compression
- Image compression
- Adaptive huffman coding in data compression
- Uniquely decodable code
- Merrval
- Compression test of ductile material
- Definition of data compression
- Lossless compression in multimedia
- Jpeg: still image data compression standard
- Jpeg still image data compression standard
- Twin city swim conference
- Griffen ford model
- Cbd inner city suburbs