1. For document similarity, the angle between two vectors is more important than the difference in their magnitude length 2. Misclassification error is not as good as either Entropy or Gini as a homogeneous measure 3. Narve Bayes makes an assumption for simplifying the computation of posterior probability 4. Unlike spla validation (eg. 80 20), k-fold cross validation allows one to test Fiery data item as an unseen data at least once. 5 Most implementations of decision tree algorithms generate a sub-optimal model 6. For a uniform distribution mean, median and mode are the same 7. PCA) and PA B) are equal undet certam properties of events A and B Triangle inequality should be satisfied by a distance meine 9 Ordinal attributes can be grouped for decision tree node split in any way possible 10 Normalizing a rector by its length to make it a unit vector) has efficiency implications