Abstract: Embodiments of the present invention relate to systems and model-free methods for perturbing neural network hardware parameters and measure the neural network response that are implemented natively within the neural network hardware and without requiring a knowledge of the internal structure of the network. Embodiments of the present invention also relate to systems and methods for configuring neural network hardware such that the network automatically performs parameter multiplexed gradient descent, which include adding a time-varying perturbation to each hardware parameter base value to modulate the cost, broadcasting the modulated cost signal to all hardware parameters, and filtering out modulations so as to extract gradient information.

Abstract: The various embodiments described herein include methods, devices, and systems for operating superconducting circuits. In one aspect, an electric circuit includes: (1) a superconductor component having a first terminal at a first end and a second terminal at a second end; (2) a gate component thermally-coupled to the superconductor component at a first location between the first terminal and the second terminal, where the gate component is thermally-coupled via a first section of the gate component; and where the gate component has a smallest width at the first section so as to focus resistive heating toward the superconductor component.

Abstract: Superconducting nanowire avalanche photodetectors (SNAPs) have using meandering nanowires to detect incident photons. When a superconducting nanowire absorbs a photon, it switches from a superconducting state to a resistive state, producing a change in voltage that can be measured across the nanowire. A SNAP may include multiple nanowires in order to increase the fill factor of the SNAP's active area and the SNAP's detection efficiency. But using multiple meandering nanowires to achieve high fill-factor in SNAPs can lead to current crowding at bends in the nanowires. This current crowding degrades SNAP performance by decreasing the switching current, which the current at which the nanowire transitions from a superconducting state to a resistive state. Fortunately, staggering the bends in the nanowires reduces current crowding, increasing the nanowire switching current, which in turn increases the SNAP dynamic range.

Abstract: Superconducting nanowire avalanche photodetectors (SNAPs) have using meandering nanowires to detect incident photons. When a superconducting nanowire absorbs a photon, it switches from a superconducting state to a resistive state, producing a change in voltage that can be measured across the nanowire. A SNAP may include multiple nanowires in order to increase the fill factor of the SNAP's active area and the SNAP's detection efficiency. But using multiple meandering nanowires to achieve high fill-factor in SNAPs can lead to current crowding at bends in the nanowires. This current crowding degrades SNAP performance by decreasing the switching current, which the current at which the nanowire transitions from a superconducting state to a resistive state. Fortunately, staggering the bends in the nanowires reduces current crowding, increasing the nanowire switching current, which in turn increases the SNAP dynamic range.