Digital electronics are based on the concept of a bit, that is either 0 or 1. By contrast, quantum bits or q-bits are delicate superpositions of 0 and 1 controlled at cryogenic temperatures. In this talk, I will introduce the concept of a probabilistic or "p-bit" that provides a bridge between the two. On the one hand, a p-bit can be viewed as a hardware accelerator for a "binary stochastic neuron" that is at the heart of a class of neural networks used in Machine Learning. On the other hand, a p-bit is like a "poor man's q-bit" that can be used to implement a class of quantum algorithms with increased versatility using existing technology that operates at room temperature. For example, commercial Magnetoresistive Random Access Memory (MRAM) technology uses stable magnets to store 0's and 1's. The same magnets can be slightly modified so that they fluctuate rapidly between 0 and 1 making them unsuitable as memory devices but ideal as hardware p-bits. However, this is just one possible implementation. Non-magnetic realizations using CMOS-based devices are also possible.
I will present several examples of device-level SPICE simulations and laboratory implementations of interconnected p-bit networks or "p-circuits" illustrating a range of applications drawn from both Machine Learning and Quantum Computing, such as, image classification, invertible logic, integer factorization and quantum annealing.
Kerem Y. Camsari graduated in 2015 with a Ph.D. in electrical engineering from Purdue University and continued as a post-doctoral researcher with the Supriyo Datta group. His PhD established a "modular approach" to connect a growing set of materials and phenomena to circuits and systems, a framework that has also been adopted by others. In his own postdoctoral work, he used this approach to establish the concept of p-bits and p-circuits as a bridge between classical and quantum circuits.