The physics of learning machines

A learning machine, like all machines, is an open system driven far from thermal equilibrium by access to a low entropy source of free energy. We discuss the connection between machines that learn, with probability error, and optimal use thermodynamic resources for both classical quantum machines. Both fixed point spiking perceptrons are discussed in context possible physical implementations. An example single photon kernel evaluation illustrates important role coherence data representation. Machine algorithms, implemented on conventional complementary metal oxide semiconductor (CMOS) devices, currently consume large amounts By focussing constraints rather than we suggest more efficient means implementing may be based switches operating at very power. Single energy efficiency might possible.