1 result
A model is PAC-learnable if it can, with high probability, learn an approximately correct hypothesis from finite samples.