Training data can be geometrically illustrated. Example: linear classifier.

Training Error

If the classifier classifies correctly 50% of the points, it misses 50% (training error \varepsilon_n(h) ) in the test data set, it’s equal to chance.

Hypothesis Space

Since each classifier represents a possible “hypothesis” regarding the data, the set of all possible classifiers could be viewed as the space of possible hypotheses.

Don’t miss these tips!

We don’t spam! Read our privacy policy for more info.

Open chat
Powered by