Susceptible Reservoir Architectures for Regime-Conditional Volatility Forecasting

Regime-Gated Residual Mixture-of-Experts for Cross-Sectional Volatility Forecasting

Financial volatility is regime dependent, yet incorporating regime information into neural networks can also destabilize training. This paper asks where such information should enter a neural cross-sectional volatility forecasting model. We study five-day realized-volatility forecasts for 1,027 U.S. equities using a rolling walk-forward evaluation framework in which information, model capacity, hyperparameter tuning, and random seeds are matched across architectures. We propose RG-ResMoE, a regime-gated residual mixture-of-experts architecture in which regime information is used only for expert routing rather than for direct forecasting. The base predictor models volatility from stock features, while a gating network uses regime state variables to route residual corrections. RG-ResMoE consistently outperforms a capacity-matched MLP in both forecasting accuracy and training stability in the main U.S. study. Similar gains are observed on an independent Japanese panel. The integration pathway is decisive: appending the same regime variables directly to the forecasting input degrades both predictive performance and training stability, whereas restricting them to the routing gate improves accuracy and Value-at-Risk calibration. Hard routing consistently underperforms soft routing. The results suggest that, in compact neural volatility forecasting models, the primary value of mixture-of-experts models lies less in increasing model capacity than in controlling how nonstationary regime information influences prediction.

Hybrid Neural-Classical Correction for Frozen Time Series Foundation Models: A Comprehensive Ablation Study on High-Frequency Stock Prediction

Foundation models for time series forecasting demonstrate impressive zero-shot generalization but often underperform on specialized domains such as high-frequency finance. We present a comprehensive study of hybrid neural-classical correction for adapting frozen TimesFM (200M parameters) to stock return prediction during the volatile opening trading hour. We compare two neural correction architectures - AttnCorrect (multi-head self-attention, approximately 471K parameters) and GatedLinear (low-rank bilinear projection with gating, approximately 49K parameters) - each augmented with Random Forest residual learning. Through systematic ablation across 10 major technology stocks (NVDA, MSFT, AAPL, GOOG, GOOGL, AMZN, META, AVGO, TSLA, NFLX) spanning 2 million data points, we reveal critical insights: (1) The hybrid neural-classical approach achieves 0.597 pooled correlation and 6.4x mean per-day correlation improvement over frozen TimesFM; (2) Classical residual learning (Random Forest) provides the largest single-component contribution, matching or exceeding the neural correction component; (3) Simpler neural architectures surprisingly outperform complex ones when classical residual learning is removed; (4) Self-attention provides the largest neural-only contribution. GatedLinear+RF achieves best overall performance with 9x fewer neural parameters than AttnCorrect+RF. We report three complementary correlation metrics - mean per-day, cross-day cumulative, and pooled - to provide a complete picture of predictive quality. Our results provide practical guidance: effective foundation model adaptation requires careful integration of neural and classical components, with classical methods playing a crucial complementary role.

Decomposition-Enhanced Network for financial time series forecasting

The extreme non-stationarity, high noise levels, and multi-timescale coupling in financial futures markets pose major challenges for time series forecasting. Existing models often struggle to disentangle localized shocks from global trends due to incompatible inductive biases. To address this issue, we propose a Decomposition-Enhanced Network (DENet). Following a divide-and-conquer paradigm, DENet adopts a multi-stream architecture: the main path extracts stable trends via moving averages and dual-path linear projections, while Auxiliary Stream I captures seasonal and local cyclical patterns using depthwise separable convolutions, and Auxiliary Stream II models high-frequency dynamics through a nonlinear autoregressive-style mapping. These components are integrated via an adaptive fusion mechanism, balancing global robustness and local structural sensitivity. Experiments on real-world futures data demonstrate that DENet outperforms a wide range of state-of-the-art benchmarks. Compared with seven state-of-the-art baselines, DENet achieves an average reduction of 10.15% in RMSE for daily forecasting on iron ore futures, and an average reduction of 22.57% in MAE under the 5-min, 12-step forecasting horizon. Furthermore, we integrate DENet’s dual-scale predictions into the R-Breaker strategy with parameter switching and dynamic position sizing. Backtesting results reveal that the annualized return of iron ore futures outperforms the baseline strategy by an average of 7.6 percentage points across five backtesting windows. Ultimately, DENet effectively bridges advanced structural modeling and actionable algorithmic trading.

A Study on Long-Horizon Stock Forecasting Failures With Deep Sequential Models

Short-term daily financial prediction is possible due to the high correlation between two consecutive daily price movements. However, long-term financial prediction through deep architectures fails due to the Markovian nature of the underlying financial dataset, resulting in correlation decay or raising temporal independence. This research uses a novel deep architecture, the KAN-RNN-Wiener framework, that integrates Kolmogorov-Arnold Networks (KAN) with physicsinformed Wiener processes to model complex non-linear financial dependencies. Although the architecture achieves superior next-day predictive accuracy over baseline deep-KAN, LSTM, and GRU models, it encounters a systemic breakdown in recursive 15-day forecasting. This paper studies how this divergence in the long-term is driven by market statistical complexities, such as non-stationarity and correlation decay. Despite advances in the architectural design proposed in this work, our observations indicate that while advanced hybrid models excel at capturing localized volatility surfaces, it fails to overcome long-term correlation decay in high-entropy environments.

Volatility forecasting with econometric and machine learning models under capacity control

Deep-MKV-TS: Path-Dependent McKean--Vlasov Control for Financial Time Series Generation

We introduce Deep-MKV-TS, a path-dependent McKean-Vlasov framework for financial scenario generation. The stochastic dynamics are chosen by matching selected path and volatility features of generated scenarios to those observed in the data. Starting from an interpretable reference model, Deep-MKV-TS preserves the reference drift and adjusts its volatility, while a regularization penalty limits unnecessary departures from the calibrated dynamics. We solve the resulting control problem using a neural, sample-based implementation of the stochastic maximum principle. We validate the method against an exactly computable oracle. On Heston and Heston-mixture models, Deep-MKV-TS substantially reduces path-dependent and volatility-related deficiencies of the reference model. In delayed-volatility experiments, the correction remains effective as the forecasting horizon increases, while direct training becomes less reliable. On held-out intraday equity-index futures, the corrected model improves conditional forecasts relative to the reference and reaches a level of performance comparable to flexible generative and historical baselines. The resulting scenarios also support greater exposure than the reference under a fixed drawdown-risk target. These results show that path-dependent McKean-Vlasov control can enrich an interpretable reference model without replacing it.