PAC learning is compression. PAC learnable, Finite VC dimensionality, and the following form of compression are fully equivalent. https://arxiv.org/abs/1610.0

PAC learning is compression. PAC learnable, Finite VC dimensionality, and the following form of compression are fully equivalent. https://arxiv.org/abs/1610.03592 https://arxiv.org/abs/1610.03592 Basically each individual neuron/perceptron just splits a space into two subspaces.