In a final project in diagnosting adhd using AI with real patients parameters. we got much higher sucsses rates in LDA than in SVM and Naive Base, we had only 100 examples in our training set.we are woundeing why lda specifically sucssed much more than the others?