Evaluating quantum circuits in the reservoir computing paradigm

arXiv:2605.01253v3 Announce Type: replace Abstract: Reservoir computing is a framework for learning temporal correlations in time series data by leveraging the intrinsic dynamics of an underlying physical system. In quantum implementations, the reservoir is commonly realised through the ergodic dynamics of nonintegrable Hamiltonian models. In our present work, we investigate the relation between reservoir performance and underlying dynamics in an analogous setting wherein a structured brickwall quantum circuit functions as the reservoir. We focus on the manner in which the choice of constituent two qubit gates influences task performance correlating the same with known results on the dynamical nature of the circuit in question. Our primary analysis involves the class of dual-unitary circuits, for which features governing the global dynamics are tunable and exactly solvable. We further study a class of circuits constructed out of non-random two qubit gates satisfying a specific solvability condition under the Cartan decomposition, for which the second moment spectrum can be determined analytically. For sufficiently large entangling power of gates, the said circuit converges to unitary designs faster than an equivalent brickwall arrangement of Haar random two qubit gates. In both cases considered, the dynamics are respectively characterised by analytically accessible exact measures of system level complexity. Our results show that reservoir performance is strongly influenced by the choice of local two-qubit operations, and that structured quantum circuits can function as effective reservoirs, where the peak performance can be accessed in relatively modest circuit depths when compared to conventional Hamiltonian reservoirs.

Abstract:Reservoir computing is a framework for learning temporal correlations in time series data by leveraging the intrinsic dynamics of an underlying physical system. In quantum implementations, the reservoir is commonly realised through the ergodic dynamics of nonintegrable Hamiltonian models. In our present work, we investigate the relation between reservoir performance and underlying dynamics in an analogous setting wherein a structured brickwall quantum circuit functions as the reservoir. We focus on the manner in which the choice of constituent two qubit gates influences task performance correlating the same with known results on the dynamical nature of the circuit in question. Our primary analysis involves the class of dual-unitary circuits, for which features governing the global dynamics are tunable and exactly solvable. We further study a class of circuits constructed out of non-random two qubit gates satisfying a specific solvability condition under the Cartan decomposition, for which the second moment spectrum can be determined analytically. For sufficiently large entangling power of gates, the said circuit converges to unitary designs faster than an equivalent brickwall arrangement of Haar random two qubit gates. In both cases considered, the dynamics are respectively characterised by analytically accessible exact measures of system level complexity. Our results show that reservoir performance is strongly influenced by the choice of local two-qubit operations, and that structured quantum circuits can function as effective reservoirs, where the peak performance can be accessed in relatively modest circuit depths when compared to conventional Hamiltonian reservoirs.

26 pages and 17 figures. Revised version includes additional classical benchmarks. Comments and feedback are welcome

arXiv:2605.01253 [quant-ph]

(or arXiv:2605.01253v3 [quant-ph] for this version)

https://doi.org/10.48550/arXiv.2605.01253 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

analiza aiqbench

Evaluating quantum circuits in the reservoir computing paradigm

Subiecte conexe