Sponsor: THALES
Context
The problem of inverting a matrix that represents group convolution and/or cross-correlation has been studied by Seth Lloyd team at MIT and at Waterloo University. The algorithm’s complexity does not depend on the sparsity of the matrix. Quantum Group Convolution is used to solve linear partial differential equations (PDEs) over domains that contain a certain symmetry (periodic PDE questions can be reformulated as cross-correlations). Achievements of polynomial speedups with respect to the dimension of the problem, while enjoying from a polylogarithmic dependence on the inverse precision.
Another Quantum Algorithm from Set Lloyd for estimating the matrix determinant based on quantum spectral sampling will be explored. The algorithm estimates the logarithm of the determinant of an n×n positive sparse matrix , exponentially faster than previously existing classical or quantum algorithms that scale linearly in n. The algorithm allows the efficient estimation of the partition function.
Objectives
Study of Quantum Group Convolution for PDE to solve:
- Elliptic equation (generalizing the Poisson equation)
Extension of study of Quantum Group Convolution for Quantum Equivariant Neural Network.
Study of Quantum estimation of the matrix determinant to: