Turing Machines and PDP Networks

Solution Preview

1. Given what you read about Turing machines and PDP networks, what is one main feature that these systems have in common? Also, what is one main difference between these two types of systems?

At a general level, Turing machines and PDP networks are similar in the sense that they are both information processors. More specifically, and more crucially, one fundamental information processing that they may share is computational power. The Universal Turing machine is known to be computationally complete, in the sense that any computation that may be carried can be carried out by such a device (according to the Church /Turing). However, there are many different examples of proof that show that PDP systems have this same power. For instance, McCullock and Pitts were able to show that one could build the machine head of a UTM from a simple network. More modern proofs have shown that PDP networks are enormously powerful pattern classifiers and functional approximators. http://www.bookrags.com/tandf/mcculloch-pitts-network-tf/

There are many differences between these two types of systems, however. In the context of information processing, one fundamental difference revolves around the structure/process distinction that is characteristic of a Turing machine. However, it is not as evident in the PDP networks. The networks both store and manipulate information, indicating that the separation between structure and process is, at the very least, extremely blurred. Another difference involves the kind of information processing that the two systems carry out. Turing machines apply rules for the specific manipulation of discrete symbols. In contrast, PDP networks appear to be subsymbolic, and are perhaps best thought ...