IVADO's Quantum Seminars Winter 2026 - Danijela Markovic, May 26th

About Danijela

Quantum neuromorphic computing with superconducting circuits

Abstract: Neuromorphic computing aims to implement neural-network-like information processing directly in hardware. Beyond conventional, isomorphic architectures that mimic the topology of neurons interconnected through synapses, non-isomorphic approaches harness more general physical systems and their intrinsic dynamics to perform computation. In this framework, input data are encoded in one of the control parameters of a physical system whose nonlinear response, governed by tunable parameters, enables complex transformations without explicitly reproducing neural network structures. Nonlinearity, which is essential for information processing, can arise from the intrinsic system dynamics, the encoding of input data, and, in quantum systems, from the measurement process through quantum back-action.

Key challenges in the field include identifying how to harness the natural evolution of a physical system for useful computation, understanding the physical origins of computational expressivity, and developing efficient training strategies adapted to quantum hardware. In this talk, I will present our recent experimental and numerical investigations of these questions using parametrically coupled bosonic modes implemented with superconducting circuits. I will discuss how encoding, control parameters, and measured observables shape the system dynamics [Dudas, npj Quant. Info., (2023); Carles, PRApplied (2026)], together with the training strategies we have explored, including model-based approaches such as backpropagation [Dudas, Sci. Rep., (2026)], and ongoing efforts towards physics-based approaches such as equilibrium propagation. These results highlight how quantum dynamics can be leveraged for neuromorphic computation.