arXiv:2611.03891
Hardware-Aware Design and Layout Selection in Quantum Reservoir Computing for Insurance Claims Forecasting
Renáta Rusnáková, Eugen Šlapak, Gaurav K. Pandey, Juraj Gazda
Abstract
Quantum reservoir computing (QRC) processes temporal information through quantum dynamics, but it remains unclear whether small QRC models can forecast real data competitively with classical models and whether their predictions survive the transfer from ideal simulation to noisy hardware. We address both questions for a six-qubit gate-based reservoir that forecasts next-day motor-insurance claim counts. Five reservoir candidates were selected and frozen using hardware-native interaction graphs, Trotterized Ising dynamics, memory and measurement-robustness criteria, circuit cost, and calibrated-noise screening. On 232 held-out test days, the ideal candidates reached root-mean-square errors (RMSEs) of 5.66-6.12 claims/day, compared with 5.36 for ridge regression. For four of the five, the difference from ridge was not significant under a 95% block bootstrap, and their accuracy was comparable with classical recurrent and reservoir baselines. On the 156-qubit IBM Kingston processor, we ran 63 evaluations in 21 paired rounds, comparing hardware-aware, random-native, and Qiskit optimization level 1 layouts. Hardware-aware placement gave lower RMSE than Qiskit in every round, with a maximum transfer distortion of 7.59 claims/day (11.84 for Qiskit), and avoided the severe transfer failures seen for some random layouts. Most of the hardware distortion appeared as a systematic forecast shift, which a short hardware probe on training inputs detected before the test runs. Longer replay windows raised circuit cost without a consistent gain in forecasting accuracy or hardware transfer. Small QRC models can thus reach competitive accuracy on this task, and hardware-aware layout selection makes their transfer to IBM Kingston more consistent.
AI Summary
The paper studies whether small quantum reservoir computers can forecast real insurance claim counts competitively with classical models, and whether hardware-aware qubit placement preserves forecasts when transferring a simulation-trained model to noisy hardware. The model uses six qubits: four input qubits reset at each step and two memory qubits, with Trotterized Ising dynamics, replayed input histories, and a frozen ridge readout trained on ideal features. Five candidates are selected using training/validation forecasting, memory, measurement robustness, circuit cost, and calibrated-noise screening. Evaluation uses 232 held-out days from one motor-insurance portfolio. Ideal QRC RMSEs are $5.66$–$6.12$ claims/day, versus $5.36$ for classical ridge; four candidates have paired block-bootstrap 95% intervals for the RMSE difference that include zero. QRC systematically underpredicts as portfolio exposure exceeds its training range. On IBM Kingston, 21 paired rounds compare hardware-aware, random-native, and Qiskit level-1 layouts, totaling 63 evaluations. Hardware-aware selection combines calibration screening with training-input QPU probes. It yields lower forecasting RMSE than Qiskit in all rounds and lower prediction-transfer distortion in 18/21. Maximum distortion is $7.59$ claims/day versus $11.84$ for Qiskit; random layouts reach $35.28$. Distortion is predominantly a systematic forecast shift, and probe distortion correlates with test distortion at $r=0.88$. Longer replay windows provide no consistent accuracy or transfer improvement. Stated limitations include one portfolio, an incomplete annual test cycle, three to five rounds per candidate on one processor, uncontrolled calibration drift, six-qubit unmitigated execution, and no second-platform hardware transfer evaluation.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.