We study a variant of nonuniform PAC learning, where the constants in the learning rate may depend on the marginal distribution, and devise a trichotomy of possible rates.

We prove that for every concept class, the optimal query complexity of agnostic active learning is strictly smaller than the sample complexity of agnostic passive learning.

We describe the geometry and combinatorics of the parameters of binary classification with starshaped polyhedral sets