STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer...

26
STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal

Transcript of STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer...

Page 1: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

STAR-Tree Spatio-Temporal Self Adjusting R-Tree

John TranDuke University

Department of Computer Science

Adviser: Pankaj K. Agarwal

Page 2: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Problem Large Moving Data Sets

Many static data structures exist, but not many account for motion, which is realistic

Page 3: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Examples of Use Geographic Information Systems

Air-Traffic Control

Protein Interactions

Traffic Patterns

Page 4: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Defining the data

Can represent data as points in Rd

For our problem: Set of data points in R2: S = {p1, p2, …, pn} Can parameterize points to pi = (xi(t), yi(t))

Piecewise differentiable velocities

Bounding boxes can be represented by 2 points

Page 5: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Queries

Query 1 – Report all points of S that lie inside rectangle R at time t

Page 6: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Queries

Query 2 – Report all points of S that lie inside rectangle R at any time between t1 and t2

Page 7: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Queries

Query 3 – Report the nearest neighbor of point in S

Page 8: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

R-Tree Bounding Box

Hierarchy All Children nodes

are bound by parents bounding box

Points are stored in leaf nodes

Page 9: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

STAR-Tree Same concept as

R-Tree Incorporate

movement into tree structure

Page 10: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Conflicts

As bounding boxes change, overlap occurs Need to adjust for these overlap conflicts

Page 11: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

QT Implementation

Page 12: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

OpenGL Implementation

Page 13: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Road Simplification Road data from US Bureau of

Census (TIGER) Paths are determined using

Dijkstra’s Shortest Path Algorithm Shapes of these paths are typically

simple but include many vertices Simplify path using Douglas-Peucker

heuristic (5 vertices max)

Page 14: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Road Simplification Simplify road network

TIGER data is not perfect Polygonal chain with vertex lists Sometimes does not match roads that

should be matched

Page 15: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Analysis of RDU RoadsV

ert

ices

wit

h n

str

eets

n streets

Page 16: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Analysis of RDU Roads

n vertices

Str

eets

wit

h n

vert

ices

Page 17: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Road Simplification

Page 18: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Protein Shape Matching

Page 19: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Problem Match two proteins based on

similarity or dissimilarity using intramolecular distance comparison

Page 20: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Data Start from PDB files

Parse to get vertex list

Page 21: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Calculating Distance Matrix Given a vertex list

Page 22: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Calculating Distance Matrix Given a vertex list

Page 23: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Defining cost

-GCTGATACTAGCT

| |||| |||||

GGGTGAT-GTAGCT

Let g(k) = +(k-1) is the cost of starting a new indel gap is the cost of continuing a gap

Page 24: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Cost Function E(i,j) = min{D(i,j-1) + , E(i,j-1) + } F(i,j) = min{D(i-1,j) + , F(i-1,j) + } D(i,j) = min{D(i-1,j-1) + (i,j),

E(i,j), F(i,j)}

Where (i,j) = normalized sum of difference distance between Ai and all the matched vertices and Bj to the corresponding matched vertices

Page 25: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Comparing identical Proteins

Page 26: STAR-Tree Spatio-Temporal Self Adjusting R-Tree John Tran Duke University Department of Computer Science Adviser: Pankaj K. Agarwal.

Test Cases