People have pointed out a couple of typo's in my slides with regard to Liu et al's paper. (1) Ujjawal below asked if the range of the y_i should be {0,1}. It should be {-1, 1}. This wasn't mentioned in Liu et al's paper but is standard for most SVM situations. (2) I had the messed up the update rule for E_i and D_i in the adaptation step I had an extra x'_t that should have been x'_i in each rule. Sorry, for the confusion. These should be easy to fix changes. Best, Chris