Exploring the feasibility of implementing hardware neural nets using memristors


June-August 2014


Me and Nicolas Merlinge, supervised by Davide Sachetto and the professor


EPFL, Microelectronic Systems Lab


The microelectronic circuit world is super excited by memristors


Can you believe I was payed to work on research concerning hardware neural nets?! Yeah, pretty weird considering I am largely a CS guy. Still, I have a EE degree in addition to a CS one, and undeniably had some VLSI experience, so worked on it I did. The problem was very loosely defined at first, and it was very interesting to converge on a goal of trying to replicate the results another group got through a software simulation with a simulation of the actual hardware that the software was emulating (very loosely, too). The details of how this was all done are in the attached summary, but suffice it to say my low-level VLSI skills were significalty tested and expanded over the summer.