Supervised Quantum Learning as Quantum Channel Simulation

We propose a formal mapping between supervised quantum learning and the information theoretic notion of channel simulation. The mapping is exploited to define a universal learning machines that can learn from quantum data.