Embedding programs into neural nets.
João Pedro Neto's NETDEF page:
http://www.di.fc.ul.pt/~jpn/netdef/ I am using J. Neto, H. Siegelmann, J. Costa, Symbolic Processing in Neural Networks, for the description of the embedding, and J. Neto, J. Costa, A. Ferreira, Merging Sub-symbolic and Symbolic Computation, for the description of higher-order synapses for learning.
The question is: how good a description of higher-order functions is possible in this framework?