Abstract
Quantum algorithms that work convincingly at small scale routinely fail at the scale where the underlying problems become interesting — often because the bottleneck limiting an algorithm at small sizes is not the one that limits it at relevant sizes.
This raises the central question the talk addresses: how to build algorithms that scale for a machine that does not yet exist and cannot be benchmarked against. planqc’s approach uses tensor networks to study problems at realistic sizes and with realistic structure entirely on classical computers, revealing where a method breaks down before we commit to it — I will show results from industrial optimisation problems. The second half of the problem is the algorithms themselves: I will present our work on making variational methods trainable at scale, which is what turns a classical insight into something a quantum computer can execute.
Biography
Martin Kiffner is Head of Quantum Solutions and Hardware Modelling at planqc, where he leads around 30 researchers working out where neutral-atom quantum computers will be genuinely useful. A theoretical physicist by training, he spent over a decade in research at Oxford and at the Centre for Quantum Technologies in Singapore and was among the first to apply tensor networks to fluid dynamics. His main interest lies at the boundary between classical and quantum methods.
