Two variants of Echo State Networks (ESN) are introduced: Fractional ESN (fESN), which incorporates fractional-differencing dynamics into the reservoir to encode long-range dependence directly, and Wavelet ESN (wESN), which extracts stable low-frequency components through wavelet smoothing before modeling them with a memory-aware reservoir.
Abstract
Accurate dengue forecasting is crucial for public health planning, but remains challenging because incidence series are often short, noisy, non-stationary, nonlinear, and often affected by long-range temporal dependence. Fractional differencing in Autoregressive Fractionally Integrated Moving Average (ARFIMA) helps balance non-stationarity and persistence, but its linear structure limits its ability to capture nonlinear dynamics. Deep neural networks can model nonlinear patterns, but usually require large training samples and do not explicitly encode statistical long memory. Echo State Networks (ESNs), a widely used reservoir computing framework, are attractive in this setting because they retain nonlinear recurrent dynamics while training only a simple readout, making them suitable for data-scarce scenarios. However, standard ESNs lack long-term memory from a time-series perspective. This study proposes a long-memory reservoir computing framework that integrates dedicated long-memory and short-memory ESN reservoirs with a ridge-regression readout. We introduce two variants: Fractional ESN (fESN), which incorporates fractional-differencing dynamics into the reservoir to encode long-range dependence directly, and Wavelet ESN (wESN), which extracts stable low-frequency components through wavelet smoothing before modeling them with a memory-aware reservoir. We establish theoretical guarantees for closed-loop reservoir dynamics, showing that standard ESNs induce short-memory processes under mild conditions, whereas the proposed long-memory reservoirs generate polynomially decaying dependence consistent with statistical long memory. Across multiple dengue datasets and forecasting horizons, fESN and wESN outperform statistical and deep learning baselines. Combining conformal prediction with fESN and wESN provides distribution-free calibrated uncertainty intervals.
TREA-Net is proposed, a Transferable Residual Epidemiological Adaptation Network for dengue forecasting under limited data that improves the corresponding backbone in 9 out of 10 settings, with statistically significant gains.
Accurate forecasts of seasonal influenza are imperative to successfully manage public health resources. However, epidemiological time series data often show significant spiky volatility along with heavy-tailed distributions that do not satisfy the normality assumption required by traditional linear models. This paper proposes a new model called Robust Laplace-ARDL, which uses a Double Exponential (Laplace) distribution instead of the standard normal distribution to accommodate heavy-tailed distributions. Using 792 weekly observations (2005–2020) and benchmarking against a Long Short-term Memory (LSTM) model, the Laplace-ARDL
(
p
=
5
)
model reduces the mean square error (MSE) by
33.5
\%
compared to the LSTM model. This paper provides empirical evidence that it is vital to solve the leptokurtic distribution in infection data for obtaining stable forecasts.
Gokul Thanigaivasan, Ratha Jeyalakshmi T, R. Mani et al.· Model Assisted Statistics an...· 0 citations
The Decomposed Recurrent Neural Network (DeRNN) is proposed, which decouples global trend modeling from local fluctuation extraction via an asymmetric dual-track architecture and exhibits superior robustness against noise and distribution shifts.
Shanyun Qian· Poster Volume 0008 The 2026...· 0 citations
This study compares ARIMA, LSTM, and temporal fusion transformer (TFT) models across three applications and shows that TFT consistently achieved superior forecasting performance and demonstrated greater robustness to increasing missingness, while k-NN generally provided the most effective imputation performance across datasets.
M. Hosseini, Mohamad Forouzanfar· Computer Science and Informa...· 0 citations
Forecasting time series over long horizons is essential for proactive decision-making in many systems. Recent research has focused on transformer-based architectures, which capture long-range dependencies in sequential data. However, several studies show that simpler linear models can outperform transformers by avoiding overfitting during training. In this context, we present NeuroFlexMLP, a deep learning model for multivariate time series forecasting tasks. NeuroFlexMLP's key distinct feature is the adaptability to the diverse complexity of real-world time series, which is achieved, from the architecture standpoint, by adding non-linear residual blocks to a first linear block. This architectural design simplifies hyperparameter optimization, leading to accurate forecasts for various time series data types regardless of the lookback or prediction horizons, outperforming state-of-the-art (SOTA) models on challenging real-world datasets. Its Multi-Layer Perceptron (MLP) design ensures high computational efficiency, making it scalable for longer input sequences than transformer-based models. We validate NeuroFlexMLP for the LEO satellite beam hopping use case, where its lightweight design enables on-board deployment, and on state-of-the art AI datasets. Across all these benchmarks, NeuroFlexMLP achieves competitive accuracy over state-of-the-art models while providing an adaptive architecture that significantly reduces computational overhead. On the LEO beam hopping task, it achieves up to 35.9% MSE reduction over Informer, which translates into up to 28% lower provisioning cost under asymmetric cost models that penalize under-allocation more heavily than over-allocation.
P. F. Pérez, Claudio Fiandrino, Marco Fiore et al.· La Main· 0 citations
Outbreaks and epidemics of infectious diseases have continuously driven the iterative development of epidemiological models. However, in reality, epidemic data often contains missing values due to delayed updates and incomplete reporting, which weakens the model's ability to characterize the transmission process and increases the difficulty of prediction. Therefore, this study proposes a loss-constrained time-varying parameter estimation neural network (L-TPENN), which directly incorporates the structural information of missing data into the objective loss function, enabling the model to handle the uncertainty caused by missing data during training. This method combines the powerful solution capabilities of Physics-Informed Neural Network under differential equation constraints with the advantages of Gated Recurrent Unit in capturing dynamic data features and handling missing data. By introducing a masking mechanism at the GRU input layer, the model can utilize the data's inherent temporal structure to execute adaptive estimation without dependence on traditional missing value imputation steps, thereby fundamentally enhancing the robustness of the estimation process. Numerical simulations show that L-TPENN achieves superior fitting performance compared to Quantile Regression Bidirectional Gated Recurrent Unit, Hybrid Grey Genetic Algorithm-based Maximum Likelihood Method and Iterated imputation estimation. Empirical analysis section, experimental results based on real pandemic data from Minnesota, demonstrate that this method can accurately fit and forecast real-world data, In furtherance of this, to make effective estimates of the time-varying parameters within the model.
Xiang-Lei Li, Jun Wang, Yue-Cai Han· Journal of Data and Dynamic...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
