| CPC G06F 30/32 (2020.01) [G06N 10/00 (2019.01)] | 20 Claims |
|
1. A method of optimizing a Gaussian transformation, the method comprising:
initializing a Gaussian transformation on n-modes parametrized by a pair of a symplectic matrix and a displacement vector;
computing an optimal Gaussian transformation using a processor that, until convergence:
evaluates a differentiable cost function that generates a scalar value indicative of quality of the Gaussian transformation;
determines a first gradient of the differentiable cost function with respect to the symplectic matrix and a second gradient of the differentiable cost function with respect to the displacement vector; and
updates the pair of the symplectic matrix and the displacement vector of the Gaussian transformation by updating the symplectic matrix using the first gradient in a geodesic optimization and updating the displacement vector using the second gradient in a gradient-based optimization;
upon convergence, extrapolating, from the pair of the symplectic matrix and the displacement vector of the optimal Gaussian transformation, circuit parameters to realize the optimal Gaussian transformation, the circuit parameters being a vector or list of parameter values; and
outputting or storing the circuit parameters.
|