Abstract
A new approach to testing cluster algorithms for invariance with respect to numbering of objects is presented. The approach is based on testing an algorithm against symmetric data. The approach allows one to avoid multiple recalculation or transformation of a similarity matrix and requires just a single use of the algorithm.
Original language | English |
---|---|
Pages (from-to) | 317-320 |
Number of pages | 4 |
Journal | Journal of Computer and Systems Sciences International |
Volume | 36 |
Issue number | 2 |
State | Published - 1997 |
Externally published | Yes |