topological reasoning between complex regions in databases with frequent updates arif khan &...

31
Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science and Engineering University of Florida Presented by: Hechen Liu

Upload: damian-reach

Post on 31-Mar-2015

218 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Topological Reasoning between Complex Regions in Databases with

Frequent Updates

Arif Khan & Markus Schneider

Department of Computer and Information Science and Engineering

University of Florida

Presented by: Hechen Liu

Page 2: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Motivation

Topological relationships are important in many applications, e.g., AI, cognitive science, and spatial databases

It is impossible to find all topological facts

It is impractical to keep all topological facts

Simple regions are not enough to represent real life scenarios

Page 3: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Complex Objects Complex regions:

− Multiple Components: faces

− Each face may have single or multiple holes

Interior: A◦

Exterior: A-

Boundary: ∂A

Page 4: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

9-Intersection Model

Page 5: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

33 Relationships of Complex Regions

[1] M. Schneider and T. Behr. Topological Relationships between Complex Spatial Objects. ACM Transactions on Database Systems, 31(1):39-81, 2006.

Page 6: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Inference

Composition

− Rx(A,B) , Ry(B,C) Rz(A , C)

− Rx o Ry Rz

− inside(A, B) o inside(B, C) inside(A,C)

Determined by the inference rules

Page 7: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Overview of the Reasoning Process

• Local Inference

− Apply inference rules

− Interpret reasoning result and identify relationship(s)

• Global Inference

− Extend the inference to N complex regions

− Binary Spatial Constraint Network (BSCN)

Page 8: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Local Inference

• Interior can characterize a complex region

• 8 possible interior-interior set relations exist between two complex regions.

A◊B: A∩ B≠ ∧ A- B≠ B- A≠

• 8*8=64 combinations possible between A and C.

Page 9: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Inference RulesConsider,

(A B⊂ ¬∂A∂B) (B∧ ∧ C⊂ ¬∂B∂C)∧A B⊂ B∧ C⊂

A C⊂

A ∩ C ≠ ∅A ∩ C = 1 (interior-interior intersection)

with the same input,

A ∩ C− = 0 (interior-exterior intersection)

A B

C

C

Page 10: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Inference Rules

Consider, A ◊ B and B ◊ C

Ao ∩ Co = unknown (interior-interior intersection)

Page 11: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Inference Rules

Page 12: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science
Page 13: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Relationship Identifying Process• If all 9 predicates are deterministic, then

inferred relationship is a single relationship.

• If there is any unknown value, then the inferred relationship is a disjunction. For example:

Page 14: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Decision Tree of the Relation Space Brute force method: 33*8=264 comparisons

Recursively divide the relationship space based on a predicate value at each level, until we reach a single relationship

− e.g.,18 relationships have false in the interior-boundary (P2) value.

33 relationships form a tree of height 6

− Deterministic values have 6 comparisons instead of 264: 97% improvement

− Indeterminate values have at most 32 comparisons: 88% improvement

Page 15: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science
Page 16: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Global Inference

Extend the reasoning process to N objects.

Binary Spatial Constraint Network (BSCN)

Page 17: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Reasoning in Dynamic Databases

Find BSCN paths

Each time a change occurs in the database, the algorithm should run

Intermediate objects are thrown out when the query is committed

Page 18: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science
Page 19: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science
Page 20: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Most Specific Relationship

The relationship which has the least number of disjunctions

− Shortest path does not guarantee most specific relationship

A CBD

E

E

A

D B

C

Page 21: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Most Specific Relationship

The relationship which has the least number of disjunctions.

− Shortest path does not guarantee most specific relationship.

overlap o overlap unknown

A CB

E

A

D B

C

Page 22: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Most Specific Relationship

The relationship which has the least number of disjunctions

− Shortest path does not guarantee most specific relationship

inside o inside inside

A CD

E

E

A

D B

C

Page 23: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Most Specific Relationship

The relationship which has the least number of disjunctions

− Shortest path does not guarantee most specific relationship

inside o disjoint disjoint

A C

E

E

A

D B

C

Page 24: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Most Specific Relationship

The relationship which has the least number of disjunctions

− Shortest path does not guarantee most specific relationship

− In fact, there is no relation between the length of the path and the most specific relationship

Page 25: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Most Specific Relationship

Solution: consider all paths and take the intersection

− Problem: number of paths is O(n!)

Interesting Facts:

− Worst case scenario when the graph is complete (then, we even do not need reasoning)

− Consider sparse graphs

Page 26: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

K-Shortest Paths

• Let us not consider all the paths. Instead, we consider k-paths

• K-shortest path algorithm: O(m+nlogn+k) [2]

• Reasoning between complex regions:

– Total complexity: O(n2 log n)

[2] D. Eppstein. Finding the k shortest paths. SIAM Journal on Computing, 28(2):652–673, 1999.

Page 27: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Simulation and Result

• Random graph

• Edges are Power Law distributed

• All edges have unit weight

• Number of paths considered: k = cn

Page 28: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Simulation and Results

Page 29: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Conclusions and Future Work

• Derived a complete set of inference rules

• Proposed BSCN and a dynamic reasoning approach

• Will introduce more robust heuristics− Weighted BSCN.

• Will extend to other data types− line-line

− line-region

Page 30: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Questions and Comments?

Please contact Mr. Arif Khan:

[email protected]

Page 31: Topological Reasoning between Complex Regions in Databases with Frequent Updates Arif Khan & Markus Schneider Department of Computer and Information Science

Thank you!