[ml] Hi

[Adam Skory]

2011/7/18 Brian Morris <cymraegish at gmail.com>:
> Another sub-discipline is "unsupervised learning" (USL). The terminology
> confused me a bit at first, they still use the term "training data" even
> though targets are unspecified.

The terminology might be clearer if you think of most ML algorithms
more generally as processes that fit parameters to maximize/minimize
certain criteria. For example in clustering, the classic case of
unsupervised learning, those criteria can be simply maximizing
similarity within groupings while minimizing it between groupings. So,
cluster assignments are still "trained" on data.

-Skory