Probabilistic Computing with Stochastic Perovskite Nickelates out in Nano Letters
Our collaborative paper with Shriram Ramanathan's group at Rutgers University is out in Nano Letters.
Congratulations Kemal Selçuk and Navid Anjum Aadit!
Excited to share our collaboration with Shriram Ramanathan's group out in Nano Letters, where we show:
(1) Energy-efficient p-bits and p-circuits can be implemented with stochastic resistors in crossbar arrays, using novel nanodevices such as hydrogen-doped perovskite nickelates.
(2) A new replica Monte Carlo algorithm we call "shared-synapse" that drastically reduces the number of multiply-accumulate operations commonly needed in a wide variety of AI applications.
Synapse sharing looks promising and may find use in alternative hardware approaches for probabilistic computing.