Memristor Crossbar Based Fuzzy Membership Functions

synaptic-plasticity Not to be confused with the other research into probabilistic computing, a new paper out on arxiv proposes an interesting method for implementing fuzzy membership functions via crossbar thresholding:

This process can simply be implemented by the memristor crossbar-based circuit shown [.fig.] Memristance of the memristors at those crosspoints which are specified by black dots in [.fig.] should be adjusted to the values written near to them in this figure. Any other memristors in the crossbar should have their highest memristance value, i.e. off R . In fact by this way, we implement the fuzzy set of [.fig.] in the antidiagonal of the crossbar through the memristance of the memristor.

And while the possibility of fuzzy processing at the hardware level is pretty interesting, they go on to propose the possibility of evolving the gates:

memristor crossbar structures like the one proposed in this paper have this potential that they can be used as a platform for implementing evolvable hardware. This is because of the fact that in these kinds of systems, variable parameters are mostly implemented at the crosspoints through the memristance of the memristors which can in return be simply modified by applying the suitable voltage or current. Note that modification of the memristance of the memristors can be done even during the execution time of the system.

Either way, the entire paper in PDF is on arxiv.org. Paper authors are Farnood Merrikh-Bayat and Saeed Bagheri Shouraki, previously of synaptic plasticity and Hebbian learning.

Taxon: | Posted in Electronics | Comment | Trackback |



Leave a Reply