Tree Based Indexing (TBI) System

TBI is a system focus on indexing multi-dimensional histograms and benefiting the Earth-Mover's Distance-based similarity search. For the B+ tree is used in TBI, our technology can be easily embedded into the real DBMS.  The technique is introduced in the following paper:

Below, the dll of our TBI system and the corresponding help document are available for download. All codes are complied with Microsoft VS 2005 in Windows XP. We have also tested the usability of the dll file under the IDE of Microsoft VS 2008 and Microsoft VS 2010. To understand the core technology of TBI system, please refer to the before-mentioned technical report. 

  • How can I get a dimensionality reduction matrix for my data?

  • Below, the dll of our TBI system and the corresponding help document are available for download. All codes are complied with Microsoft VS 2005 in Windows XP. We have also tested the usability of the dll file under the IDE of Microsoft VS 2008 and Microsoft VS 2010. To understand the core technology of TBI system, please refer to the before-mentioned technical report. 

    Efficient EMD-based similarity search in multimedia databases via flexible dimensionality reduction, SIGMOD 08: 199-212.

  • Why the TBI search results do not match the linear scan results?

  • Check if you have used the same ground distance (e.g., Euclidean distance) during the process of feasible solution file generation and the process of TBI running. That is to say, if you use the Euclidean distance-based feasible solution file, you have to set the ground distance as Euclidean distance in the file parameter.h. Otherwise, some results will get lost. 

  • What is the largest histogram dimensionality and database cardinality TBI system can support?
  • The largest histogram dimensionality and database cardinality the recent version of TBI can support are 250 and 500,000.  If you need a larger histogram dimensionality of database cardinality to support your application, please don't hesitate to contact us.

If you encounter any problems with the algorithms or the implementations of TBI, please make sure you have read the technical report above,the user guide, and the Q&A first. If your answer still can not be solved well, please feel free to contact us (xujia AT ise.neu.edu.cn). We also get quite interested in your obtained results by using the TBI system. 

Northeastern University |Shenyang| China|110004
Copyright 2010 All rights reserved.
 
     
ҳģ