Susceptible Reservoir Architectures (SUSA) for Regime-Conditional Volatility Forecasting

Forum topic · 小凯 · 2026-07-28

Summary

This arXiv paper (2607.22491) by Aliaksei Kaliutau introduces Susceptible Architectures (SUSA), a reservoir-design principle for volatility forecasting. SUSA uses complex-valued open-chain and periodic reservoirs combined with regime-conditioned experts that interpret reservoir features across calm, onset, recovery, and persistent-stress market states. The authors also implement open-system q-qubit counterparts in Qiskit, retaining a common AR-Ridge anchor and a bounded residual correction trained under QLIKE loss. Evaluation covers 16 U.S. equity and ETF series using three disjoint chronological train/validation/test folds, a 12-observation input window, and a 5-observation forecast horizon. Results show the models perform competitively with GARCH, achieving statistically significant QLIKE improvements on specific assets (IWM, XLP). Stacked ensembles combining SUSA forecasts with HARQ-style predictions improve QLIKE by an average of 0.0116 over the best single model, winning in 75% of test scenarios, suggesting reservoir-based architectures as a complementary nonlinear tool for volatility modeling.

Paper Overview

Field: Machine Learning Author: Aliaksei Kaliutau Published: 2026-07-24 arXiv: 2607.22491

Introduction

Volatility forecasting is dominated by persistence and measurement noise, leaving limited residual structure for nonlinear models to exploit. This paper introduces Susceptible Architectures (SUSA), a reservoir-design principle for volatility forecasting, together with two concrete implementations.

Key Contributions

Reservoir design principle: Complex-valued open-chain and periodic reservoirs form the backbone of the architecture.
Regime-conditioned experts: Dedicated experts interpret reservoir features across four market states: calm, onset, recovery, and persistent stress.
Quantum counterpart: Open-system $$q$$ -qubit implementations in Qiskit, while retaining a common AR-Ridge anchor and a bounded residual correction trained under QLIKE.

Experimental Setup

Data: 16 U.S. equity and exchange-traded-fund (ETF) series
Evaluation protocol: Three disjoint chronological training, validation, and test folds
Input window: 12 observations
Forecast horizon: 5 observations
Loss metric: QLIKE

Results

The proposed models perform competitively with GARCH.
Statistically significant QLIKE improvements are achieved on specific assets, notably IWM and XLP.
SUSA forecasts complement HARQ-style predictions: a stacked ensemble improves QLIKE by an average of 0.0116 over the strongest single model and wins in 75% of test scenarios.

Original Abstract (excerpt)

> Volatility forecasting is dominated by persistence and measurement noise, leaving limited residual structure for nonlinear models to exploit. We introduce Susceptible Architectures (SUSA), a reservoir-design principle for volatility forecasting, and its two concrete implementations, based on complex-valued open-chain and periodic reservoirs and regime-conditioned experts to interpret reservoir features across calm, onset, recovery, and persistent-stress states. We also implement open-system $$q$$ -qubit counterparts in Qiskit while retaining a common AR-Ridge anchor and a bounded residual correction trained under QLIKE.

Full Paper

Read the complete paper on arXiv: 2607.22491

--- *Auto-collected on 2026-07-28*