US 12,033,030 B2
Apparatus and methods for quantum computing and machine learning
Nathan Killoran, Toronto (CA); Thomas R. Bromley, Toronto (CA); Juan Miguel Arrazola, Toronto (CA); Maria Schuld, Durban (ZA); and Nicolas Quesada, Toronto (CA)
Assigned to Xanadu Quantum Technologies Inc., Toronto (CA)
Filed by Xanadu Quantum Technologies Inc., Toronto (CA)
Filed on Jan. 23, 2023, as Appl. No. 18/100,290.
Application 18/100,290 is a continuation of application No. 16/444,624, filed on Jun. 18, 2019, granted, now 11,562,279.
Prior Publication US 2023/0153668 A1, May 18, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06N 10/00 (2022.01); G06N 3/044 (2023.01); G06N 3/067 (2006.01); G06N 20/00 (2019.01)
CPC G06N 10/00 (2019.01) [G06N 3/044 (2023.01); G06N 3/067 (2013.01); G06N 20/00 (2019.01)] 19 Claims
OG exemplary drawing
 
1. An apparatus, comprising:
a plurality of processing layers coupled in series, each processing layer in the plurality of processing layers including:
a Gaussian unit configured to perform a linear transformation on an input signal including a plurality of optical modes, the Gaussian unit including:
a network of interconnected beamsplitters and phase shifters, a first portion of the network of interconnected beamsplitters and phase shifters being configured to form a first interferometer to apply one of a phase-less Gaussian transformation or a phase-sensitive Gaussian transformation to the plurality of optical modes, and a second portion of the network of interconnected beamsplitters and phase shifters being configured to form a second interferometer to apply a phase-less transformation to the plurality of optical modes; and
a plurality of squeezers operatively coupled to the network of interconnected beamsplitters and phase shifters, each squeezer from the plurality of squeezers configured to apply one of a position-only squeezing or squeezing along an arbitrary axis in a position-momentum plane to one optical mode from the plurality of optical modes, and
a plurality of nonlinear gates operatively coupled to the Gaussian unit and configured to apply an arbitrary nonlinear transformation to the plurality of optical modes; and
a controller operatively coupled to the plurality of processing layers and configured to change a setting of the plurality of processing layers based on an output signal from at least one processing layer from the plurality of processing layers, to vary the linear transformation and the arbitrary nonlinear transformation to switch the apparatus among a first mode to implement a classical neural network, a second mode to implement a quantum neural network, and a third mode to implement a hybrid neural network.