Speaker
Description
Quantum reservoir computing (QRC) is an appealing quantum machine learning model: the quantum system acts as a fixed dynamical feature map and only a linear readout is trained, so learning reduces to a convex problem that avoids the barren plateaus and costly variational loops of quantum neural networks, while remaining well suited to near-term hardware. Yet QRC has been studied almost exclusively on univariate time series, and whether genuinely quantum resources help when processing multivariate data is unknown. We establish a systematic framework for multivariate QRC: we propose three multivariate encoding schemes, introduce the mixing capacity to quantify how effectively a reservoir combines independent data streams, and benchmark two physically distinct quantum reservoirs, a discrete-variable (DV) transverse-field Ising system and a continuous-variable (CV) network of coupled harmonic oscillators. We find that the optimal encoding is strongly system- and task-dependent and, crucially, that peak performance coincides with regimes exhibiting non-classical resources, namely entanglement (DV) and squeezing (CV), linking quantum resources to multivariate processing power.
| Topic | Encoding and processing of data in quantum systems |
|---|---|
| Additional Topic | Hybrid quantum-classical approaches for learning and optimization |