Supervised Learning
A classifier training [see page 7, process] that learns, using labelled sample data, how to classify unlabelled data.
Samples are supplied as: \[ \{({x}^{1}, {t}^{1}), ({x}^{2}, {t}^{2}), \ldots, ({x}^{p}, {t}^{p}) \} \] With:
- \( {x}^{i} \) being the sample \( i \)
- \( {t}^{i} \) being the label of a sample \( i \).
The main [see page 8, goal] in supervised learning is using the known classes of samples to tune the classifier (learn the weights) that minimise the error/cost function. \[ E = F(\text{input}, \text{output}; w) \] Use the input and output to produce a new set of weights.