TEAM QEDI
EPFL Quantum Hackathon • Quandela Challenge
Predicting the next H day volatility surface topologies.
Reconstructing missing Option Price nodes in the 2D surface grid.
Inspired by Li et al., "QRC for Realized Volatility Forecasting" (2024), we adapted their qubit-based Hamiltonian evolution into a purely Photonic Quantum Reservoir powered by MerLin.
Instead of mapping data to all spatial modes simultaneously, our temporal array uses 5 input modes and 3 dedicated memory modes. The memory modes are left unencoded, continuously accumulating historical state contexts through serial phase mixing.
We sample the evolving physical system at multiple structural post-processing depths. These Virtual Nodes emulate capturing chronological measurement sub-intervals (δτ), massively expanding our temporal feature dimensionality without adding physical photon bounds.
Instead of measuring impossibly vast raw Fock states, we group the probability vectors via LexGrouping across 3 random seeds × 3 virtual depths. By utilizing this Hybrid Photonic Temporal QRC architecture, we massively outperformed our standard formulation!
Extracting the most prominent Non-linear Mutual Information quantum channels, an L2 regularized Ridge projection predicts the consecutive future states along the latent PCA coordinate axis.
By executing the Hybrid Photonic Temporal QRC pipeline (Li et al., 2024), we successfully beat our underlying standard QRC framework and left classical baseline LSTM networks and quantum regression tools absolutely obsolete.
RMSE Error
RMSE Error
RMSE Error
RMSE Error
RMSE Error
RMSE Error
Performance loss values compared across classical, hybrid, and pure photonic MerLin approaches.
Interactive Error Grid: (Predicted − Actual) across all 224 features.
Tenor vs Maturity Average MAE Grid over the 6-Day Inference Horizon.
Teaching photons to predict the market so we can finally sleep.
