Weak computationalism: all models must be computational Strong c.: things are computational Cognitive c.: the mind is computational Universal c.: everything is computational Digital computation: mappings between [finite] ints Quantum c.: between complex Hyper c.: more than int
I don't see the relation. Tensorflow is a library that supports particular classes of function approximators, but not all the relevant ones. Ultimately, you want to learn arbitrary algorithms; even simple loops are impossible in feedforward networks, and need GRUs etc.