digital-quantum-reservoir-computing

name	digital-quantum-reservoir-computing
description	Digital quantum reservoir computing (QRC) framework for time series forecasting on near-term quantum devices. Uses parametrized four-qubit reservoirs with partial measurement and reset, encoding temporal data in rotation angles. Training restricted to classical Ridge-regression readout. Use when: quantum reservoir computing, time series forecasting, near-term quantum devices, ATM cash demand prediction, quantum ML for financial data.
metadata	{"arxiv_id":"2606.04686","published":"2026-06-03","authors":"Chiara Vercellino, Giacomo Vitali, Valeria Zaffaroni, Francesca Cibrario, Emanuele Dri, Paolo Viviani, Olivier Terzo, Davide Corbelletto","tags":["quantum-ml","reservoir-computing","time-series","forecasting","near-term-quantum"]}

Digital Quantum Reservoir Computing for Time Series

Core Concept

Digital QRC uses parametrized four-qubit reservoirs with fixed structure exploiting partial measurement and reset. Temporal data encoded in rotation angles; training restricted to classical Ridge-regression readout. Systematically analyzed circuit ansatz, reservoir memory, measurement-derived observables, and execution backend impact on forecasting performance.

Architecture

Quantum Reservoir

• 4-qubit parametrized reservoir with fixed structure
• Partial measurement and reset between time steps
• Rotation angle encoding of temporal input data
• No trainable quantum parameters — only classical readout

Classical Readout

• Ridge regression on quantum measurement outcomes
• Observable selection from measurement-derived operators
• Hyperparameter tuning only at classical level

Key Findings

• QRC does NOT outperform classical Prophet benchmark in MAE/NMSE
• More competitive results in Dynamic Time Warping metric
• Partial ability to capture temporal structure despite metric limitations
• Validated on noiseless simulation, noise-aware emulation, and real IQM Spark quantum processor

Workflow

Step 1: Data Encoding

1. Convert time series to rotation angles via encoding function
2. Apply encoded rotations to 4-qubit reservoir circuit

Step 2: Reservoir Evolution

1. Run parametrized reservoir circuit
2. Partial measurement of qubits
3. Reset measured qubits to |0⟩
4. Repeat for each time step

Step 3: Classical Training

1. Extract measurement-derived observables
2. Train Ridge regression on reservoir outputs
3. Validate on held-out time series segments

Step 4: Evaluation

1. MAE, NMSE for point prediction accuracy
2. Dynamic Time Warping for temporal structure capture
3. Compare against classical baselines

Hardware Considerations

• Tested on IQM Spark quantum processor
• Noise-aware emulation for intermediate validation
• Near-term device limitations affect performance

Activation Keywords

• quantum reservoir computing
• digital quantum reservoir
• quantum time series forecasting
• QRC financial prediction
• quantum ML time series
• near-term quantum machine learning