Memcomputing: from abstract machines to practical realizations.
Memcomputing is a computing paradigm recently introduced to describe a new class of machines based on (non von Neumann) architectures formed by interconnected memory cells capable to change their state according to the stimuli from the collective state of the network.
The power of these machines can be increased exponentially if we embed in the network some extra information related to the problem aimed to solve.
We call this feature information overhead. By taking advantage of this embedded information we can exponentially reduce the complexity of many problems such as the Non deterministic Polynomial (NP) ones. In fact, they can be in principle solved by
a memcomputing machine with only polynomial resources (time, space, energy).
It will also be presented a practical realization of these machines using what we named self-organizing logic gates, i.e., logic gates that can accept inputs from all terminals, including the conventional output terminals.
Moreover, it will be showed how to use them to solve NP problems with polynomial resources.
Speaker: FABIO LORENZO TRAVERSA
(3 MB)