SGT

2

1. Binary classifier to choose whether to display white or black text on a given background colour. The feature set is the three colour channels for the background, RGB. This can be used to increase accessibility of a website by automatically adapting text to have the most visibility.
2. ..
3. ..
4. Predict carbon emission levels using electricity used as a feature.

3

1. There is no training. Usually, training classifier involves processing the training set in some way. In a nearest-neighbour classifier, we use the training set raw.
2. Method on slides:
   • $k$ elements of training set that are nearest to new example are identified
   • number of these in each class $C_1,C_2,\dots$ are counted, call these $n_1,n_2,\dots$
   • probability of new example being in each class is then calculated as follows: $p(C_1)=\frac{n_1}{k}$ $p(C_2)=\frac{n_2}{k}$ and so on for other classes.

4