Conference Thoughts
some things I found interesting to mention.
Educational
- Traditional measures for association rules
- support: P(XY)
- confidence: P(X|Y)
- lift
- Interpretation: assigns object to terms (analogous to a function);
I: T -> P(Obj) I(Crete) = {hotel1, hotel2}
- OWL-DL: open world assumption => negation (if not explicitly specified for the object) does not deliver any result); compare: you cannot prove that a white swan does not exists.
Statistical Methods
- Kullback-Leibler Divergence between two probability functions
- Goodman Kriskal tau-measure (for categorical data): \[ \tau = (E_D - E_D|I) / (E_D) \] where $$I$$, $$D$$ are concepts