jeremykun.com
A problem that is not (properly) PAC-learnable
In a previous post we introduced a learning model called Probably Approximately Correct (PAC). We saw an example of a concept class that was easy to learn: intervals on the real line (and more gene…