Data for Evaluating Learning
in
Valid Experiments

U of T

--------------------------------------------------------------------

Note: The Delve project is in abeyance at the moment, and the software and archive here are not actively maintained. It appears that the software does not work with the most recent versions of tcl. A new version of the Delve software has been mostly written (in C this time), but requires more testing before it is released. This may happen by the end of the summer in 2003.

Delve is a standardised environment designed to evaluate the performance of methods that learn relationships based primarily on empirical data. Delve makes it possible for users to compare their learning methods with other methods on many datasets. The Delve learning methods and evaluation procedures are well documented, such that meaningful comparisons can be made.

Delve is evolving: we encourage you to submit datasets and new learning methods to the delve archive.

Delve consists of:

  1. A software environment, which allows you to manipulate datasets and do statistical analysis of method performance. The users manual is available in postscript. You must install this software to gain maximum benefit from delve. Download thelatest version of the software. Please check for the latest patches.
  2. Datasets; a number of datasets for regression and classification are available. The list is growing, please let us know if you have datasets that could be suitable for delve. Experienced Delve users may wish to immediately view the dataset summary table.
  3. Learning methods. Delve also serves as a repository of learning methods. In the archive you will find precise descriptions of methods, software implementations and complete results obtained by running the methods on the datasets. Researchers are encouraged to submit their learning methods to delve.
Delve is copyrighted, and you should read the copyright notice before using it.

There is a electronic mailing list for users of delve; to join the list, send a request to delve-request@cs.utoronto.ca. Send any comments on Delve to delve@cs.utoronto.ca.

You can visit the home pages of the members of the Delve development group, or view acknowledgements.


Last modified: Thu Nov 27 12:52:32 EST
Comments and questions to: delve@cs.toronto.edu
Copyright