CS1674: Homework 6 - Written

1. [4 pts] Briefly describe what we do at training and test time, when learning a prediction function for classification.
2. [3 pts] In 1-nearest-neighbor, we just copy the label of the closest training sample. What additional step do we need to perform for k-nearest-neighbors, where k > 1?
3. [3 pts] Why do we want a large margin between positive and negative samples, in support vector machines?