An integrated optical modulator operating at cryogenic temperatures.

Photonic integrated circuits (PICs) operating at cryogenic temperatures are fundamental building blocks required to achieve scalable quantum computing and cryogenic computing technologies1,2. Silicon PICs have matured for room-temperature applications, but their cryogenic performance is limited by the absence of efficient low-temperature electro-optic modulation. Here we demonstrate electro-optic switching and modulation from room temperature down to 4 K by using the Pockels effect in integrated barium titanate (BaTiO3) devices3. We investigate the temperature dependence of the nonlinear optical properties of BaTiO3, showing an effective Pockels coefficient of 200 pm V-1 at 4 K. The fabricated devices show an electro-optic bandwidth of 30 GHz, ultralow-power tuning that is 109 times more efficient than thermal tuning, and high-speed data modulation at 20 Gbps. Our results demonstrate a missing component for cryogenic PICs, removing major roadblocks for the realization of cryogenic-compatible systems in the field of quantum computing, supercomputing and sensing, and for interfacing those systems with instrumentation at room temperature.

Neuromorphic Systems Design by Matching Inductive Biases to Hardware Constraints.

Neuromorphic systems are designed with careful consideration of the physical properties of the computational substrate they use. Neuromorphic engineers often exploit physical phenomena to directly implement a desired functionality, enabled by "the isomorphism between physical processes in different media" (Douglas et al., 1995). This bottom-up design methodology could be described as matching computational primitives to physical phenomena. In this paper, we propose a top-down counterpart to the bottom-up approach to neuromorphic design. Our top-down approach, termed "bias matching," is to match the inductive biases required in a learning system to the hardware constraints of its implementation; a well-known example is enforcing translation equivariance in a neural network by tying weights (replacing vector-matrix multiplications with convolutions), which reduces memory requirements. We give numerous examples from the literature and explain how they can be understood from this perspective. Furthermore, we propose novel network designs based on this approach in the context of collaborative filtering. Our simulation results underline our central conclusions: additional hardware constraints can improve the predictions of a Machine Learning system, and understanding the inductive biases that underlie these performance gains can be useful in finding applications for a given constraint.