Back to Search
Start Over
Differentiation of Linear Optical Circuits
- Publication Year :
- 2024
-
Abstract
- Experimental setups based on linear optical circuits and single photon sources offer a promising platform for near-term quantum machine learning. However, current applications are all based on support vector machines and gradient-free optimization methods. Differentiating an optical circuit over a phase parameter poses difficulty because it results in an operator on the Fock space which is not unitary. In this paper, we show that the derivative of the expectation values of a linear optical circuit can be computed by sampling from a larger circuit, using one additional photon. In order to express the derivative in terms of expectation values, we develop a circuit extraction procedure based on unitary dilation. We end by showing that the full gradient of a universal programmable interferometer can be estimated using polynomially many queries to a boson sampling device. This is in contrast to the qubit setting, where exponentially many parameters are needed to cover the space of unitaries. Our algorithm enables applications of photonic technologies to machine learning, quantum chemistry and optimization, powered by gradient descent.
- Subjects :
- Quantum Physics
Subjects
Details
- Database :
- arXiv
- Publication Type :
- Report
- Accession number :
- edsarx.2401.07997
- Document Type :
- Working Paper