Colloquium: Gaweł Kuś (ASM)
17 december 2019 15:00 - Locatie: Lecture Room D, Faculty of Aerospace Engineering, Kluyverweg 1, Delft
Quantum Gaussian Processes for Data-Driven Design of Metamaterials
The data-driven approach shows great potential for computational design and optimization of materials. Recently, a framework relying on Gaussian processes was employed for the computational design of metamaterial that achieves super-compressibility despite being manufactured out of intrinsically brittle materials. While showing superior predictive performance, the method, however, suffers from severe scaling issues, which limit its application to problems with only few design parameters.
This research explores surpassing those classical limitations with quantum computing, specifically, by replacing the machine learning step with an exponentially faster quantum algorithm for Gaussian processes. For this purpose, the quantum algorithm was implemented and applied (as a proof-of-concept) within the data-driven framework to a simple case of metamaterial design. Simulation and numerical analysis of the algorithm additionally exposed a new connection between the quantum Gaussian processes and the classical sparse approximations.