Expressing and Optimizing the Similaritybased Queries In SQL

1
Expressing and Optimizingthe Similarity-based
QueriesIn SQL

• Part3 Research topic

2
Similarity-based Search
Review general topic
Topic 1
Topic 2
Query Optimization

• Domain specific questions that ask for the
  nearest/near neighbors from a data set
• e.g.
• (GIS) What are the 3 nearest gas stations to GMU
  Fairfax campus?
• (Image) 4D MRI tracking which of the previous
  tumor images is closest to current one?

• One of database techniques
• Goal automatically determine the best execution
  plan
• How to perform each operator and what is the
  ordering of these operators?
• Support standard operators and built-in
  functions , -, , /, max, min, etc.
• Not supported
• User-Defined-Functions

GIS geographic information system
3
Similarity-based Search
Review general topic
Database Systems
Data
UDT
Search functions
UDF
Query optimization?
4
Two problems
Research topic problems

• Model the similarity-based searches
• Need a general form to represent similarity-based
  searches from different domains
• Optimize an SQL with these searches
• How to most efficiently answer the SQL query that
  contains similarity-based searches?

5
Research topic modeling the similarity-based
search
Search examples

• Q1 Which are 3-nearest gas stations to GMU?
• Q2 Which are 3-nearest Shell gas stations to
  GMU?
• Q3 Which Shell stations are among the 3-nearest
  gas stations to GMU?
• We need an abstract form to represent all such
  queries.
• ?
• Q Which elements in a set are k-nearest
  neighbors of a query data in a domain?

6
Research topic modeling the similarity-based
search
Model of the search
Default data domain
Given a default data domain, the similarity-based
search to be studied is modeled as
where the result set Elements belongs to Set.
This is the basic hypothesis in this research.
Elements
Set
query data
an element