A matrix of meh-ristors
"“Probabilistic programming languages (PPLs) solve these problems by marrying probability with the representational power of programming languages. A probabilistic program is a mix of ordinary deterministic computation and randomly sampled values.”"
I'm fairly ignorant in this area (except for what I read on the El Reg, oy) but it strikes me that this approach notes the 'apparent randomness' of a real neural network, and kinda 'fakes' that randomness in a purely logical digital realm by shoehorning it in. It's interesting stuff, but I don't see how this approach can really replicate a NN's ability to adapt in a very complex way to inputs.
As I say tho, I'm not an expert here.