※ 引述《rafan (@[email protected]"""")》之铭言：
: ※ 引述《hometoofar (家太远了)》之铭言：
: Homework 1 is using k-nearest neighbor method, not SVM.
: Just forget it.The dataset is a multilabel one, i.e.,
: one instance can have more than one target value (class
: labels).
: Since kNN is original designed for one target value,
: it is up to you that how to apply it on multilabel
: dataset.
Since we are using kNN here, does that mean we don't need to train it? We just
use the training data and kNN to predict what class label should a test entry
have and compare it to the true class label? Then we compare different values
of k and see which k would give the best result?