Why
was SKICAT Effective?
-
Problem of feature measurement was performed
completely by user.
-
Astronomers knew what features to measure
but not how to use them to classify ski objects.
-
Accurate classification needed at least
8 of 40 dimensions simultaneously
-
Separating stars from galaxies at faint
levels was a significant open problem:
-
scatter plots could not overcome high
dimensionality
-
models were complex, and relied on piecewise
approximation over small partition regions.
