Shihao Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Shihao Yang has a total of 141 co-authors. Here are the top co-authors by number of shared publications:
Simin Ma
H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology , Atlanta, GA 30332
5 shared publications- Jiecheng Lu
Georgia Institute of Technology
4 shared publications - Samuel C Kou
Department of Statistics Harvard University 1 Oxford Street 02138 Cambridge MA United States
4 shared publications - Samuel W. K. Wong
Department of Statistics and Actuarial Science, University of Waterloo
4 shared publications - Simin Ma
H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332
4 shared publications Jiecheng Lu
Georgia Institute of Technology, Atlanta, GA, USA
3 shared publications
Latest Publications
- Open Access
Multi-season evaluation and analysis of categorical trend forecasts of influenza hospital admissions in the United States
Forecasting influenza hospitalizations informs public health preparedness, yet questions remain about which types of forecasts best guide action. We evaluate categorical trend forecasts, which communicate probabilities of upcoming increases or decreases in epidemic trajectories, submitted to CDC’s FluSight Forecasting Challenge between Fall-2024 and Spring-2026. Teams submitted probability distributions over five categories describing direction and magnitude of week-over-week changes in laboratory-confirmed influenza hospital admissions. We assessed performance using Ranked Probability Skill Score, Brier Skill Score, and measures of forecast-observation agreement. Most models outperformed an equal-probability baseline; the FluSight ensemble ranked among the top three in the 2024–25 and 2025–26 seasons. Forecasts were most accurate during stable periods and least during periods of rapid change, with most models underestimating observed trends. Conclusions were robust to choice of scoring metric and reference model. These results support categorical trend ensembles as an approach to communicating infectious disease forecasts that may inform public health decision-making.
Open AccessMulti-season evaluation and analysis of categorical trend forecasts of influenza hospital admissions in the United States
Forecasting influenza hospitalizations informs public health preparedness, yet questions remain about which types of forecasts best guide action. We evaluate categorical trend forecasts, which communicate probabilities of upcoming increases or decreases in epidemic trajectories, submitted to CDC’s FluSight Forecasting Challenge between Fall-2024 and Spring-2026. Teams submitted probability distributions over five categories describing direction and magnitude of week-over-week changes in laboratory-confirmed influenza hospital admissions. We assessed performance using Ranked Probability Skill Score, Brier Skill Score, and measures of forecast-observation agreement. Most models outperformed an equal-probability baseline; the FluSight ensemble ranked among the top three in the 2024–25 and 2025–26 seasons. Forecasts were most accurate during stable periods and least during periods of rapid change, with most models underestimating observed trends. Conclusions were robust to choice of scoring metric and reference model. These results support categorical trend ensembles as an approach to communicating infectious disease forecasts that may inform public health decision-making.
- Open Access
- 15 June 2026
- Published by Taylor & Francis in The American Statistician
- Vol. 80 (4) , 491-506
Are Statistical Methods Obsolete in the Era of Deep Learning? A Study of ODE Inverse Problems
- 15 June 2026
- Published by Taylor & Francis in The American Statistician
- Vol. 80 (4) , 491-506
Open AccessAre Statistical Methods Obsolete in the Era of Deep Learning? A Study of ODE Inverse Problems
- Open Access
- 19 May 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
A real-time early warning system to anticipate respiratory disease outbreaks using transfer learning
Respiratory disease outbreaks burden American healthcare systems with over one million hospitalizations annually, yet current surveillance systems lag 1-2 weeks behind real-time conditions, preventing timely intervention. We present a machine learning early warning system that combines Google search trends with traditional epidemiological data using ensemble voting algorithms to predict outbreak timing across multiple respiratory pathogens. Unlike prior digital surveillance systems focused on retrospective evaluation or single-pathogen settings, this work presents a unified, prospectively deployed early warning framework that detects both outbreak onsets and peaks across multiple respiratory pathogens at the state level in real time. The system applies anomaly detection and transfer learning to monitor syndromic influenza-like illnesses, and hospitalizations caused by respiratory syncytial virus or influenza, simultaneously, across all 50 states. During operational real-time deployment from August 2024 through the 2024-2025 season, the system detects 98.0% of outbreak onsets and 97.0% of outbreak peaks, with average lead times of approximately 5 and 2 weeks, respectively, and positive predictive values exceeding 82%. This framework transforms reactive public health responses into proactive epidemic preparedness by reducing historical timing uncertainty from 10-20 weeks to consistent 2-6 week prediction windows, providing a scalable approach for monitoring both seasonal outbreaks and emerging respiratory threats.
- 19 May 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessA real-time early warning system to anticipate respiratory disease outbreaks using transfer learning
Respiratory disease outbreaks burden American healthcare systems with over one million hospitalizations annually, yet current surveillance systems lag 1-2 weeks behind real-time conditions, preventing timely intervention. We present a machine learning early warning system that combines Google search trends with traditional epidemiological data using ensemble voting algorithms to predict outbreak timing across multiple respiratory pathogens. Unlike prior digital surveillance systems focused on retrospective evaluation or single-pathogen settings, this work presents a unified, prospectively deployed early warning framework that detects both outbreak onsets and peaks across multiple respiratory pathogens at the state level in real time. The system applies anomaly detection and transfer learning to monitor syndromic influenza-like illnesses, and hospitalizations caused by respiratory syncytial virus or influenza, simultaneously, across all 50 states. During operational real-time deployment from August 2024 through the 2024-2025 season, the system detects 98.0% of outbreak onsets and 97.0% of outbreak peaks, with average lead times of approximately 5 and 2 weeks, respectively, and positive predictive values exceeding 82%. This framework transforms reactive public health responses into proactive epidemic preparedness by reducing historical timing uncertainty from 10-20 weeks to consistent 2-6 week prediction windows, providing a scalable approach for monitoring both seasonal outbreaks and emerging respiratory threats.
- Top Cited Papers
- 31 March 2026
- Published by Elsevier in International Journal of Forecasting
- Vol. 42 (3) , 1104-1122
Restoring the forecasting power of Google Trends with statistical preprocessing
- 31 March 2026
- Published by Elsevier in International Journal of Forecasting
- Vol. 42 (3) , 1104-1122
Top Cited PapersRestoring the forecasting power of Google Trends with statistical preprocessing
- Open Access
HyperMLP: An Integrated Perspective for Sequence Modeling
Self-attention is often viewed as probabilistic query-key lookup, motivating designs that preserve normalized attention scores and fixed positional semantics. We advocate a simpler and more unified perspective: an autoregressive attention head can be viewed as a dynamic two-layer MLP whose weights are instantiated from the context history. From this view, attention scores form an ever-growing hidden representation, and standard MLP activations such as ReLU or GLU naturally implement input-conditioned selection over a context-dependent memory pool rather than a probability distribution. Based on this formulation, we introduce HyperMLP and HyperGLU, which learn dynamic mixing in both feature space and sequence space, using a reverse-offset (lag) layout to align temporal mixing with autoregressive semantics. We provide theoretical characterizations of the expressivity and implications of this structure, and empirically show that HyperMLP/HyperGLU consistently outperform strong softmax-attention baselines under matched parameter budgets.
Open AccessHyperMLP: An Integrated Perspective for Sequence Modeling
Self-attention is often viewed as probabilistic query-key lookup, motivating designs that preserve normalized attention scores and fixed positional semantics. We advocate a simpler and more unified perspective: an autoregressive attention head can be viewed as a dynamic two-layer MLP whose weights are instantiated from the context history. From this view, attention scores form an ever-growing hidden representation, and standard MLP activations such as ReLU or GLU naturally implement input-conditioned selection over a context-dependent memory pool rather than a probability distribution. Based on this formulation, we introduce HyperMLP and HyperGLU, which learn dynamic mixing in both feature space and sequence space, using a reverse-offset (lag) layout to align temporal mixing with autoregressive semantics. We provide theoretical characterizations of the expressivity and implications of this structure, and empirically show that HyperMLP/HyperGLU consistently outperform strong softmax-attention baselines under matched parameter budgets.
- Open Access
StretchTime: Adaptive Time Series Forecasting via Symplectic Attention
Transformer architectures have established strong baselines in time series forecasting, yet they typically rely on positional encodings that assume uniform, index-based temporal progression. However, real-world systems, from shifting financial cycles to elastic biological rhythms, frequently exhibit "time-warped" dynamics where the effective flow of time decouples from the sampling index. In this work, we first formalize this misalignment and prove that rotary position embedding (RoPE) is mathematically incapable of representing non-affine temporal warping. To address this, we propose Symplectic Positional Embeddings (SyPE), a learnable encoding framework derived from Hamiltonian mechanics. SyPE strictly generalizes RoPE by extending the rotation group $\mathrm{SO}(2)$ to the symplectic group $\mathrm{Sp}(2,\mathbb{R})$, modulated by a novel input-dependent adaptive warp module. By allowing the attention mechanism to adaptively dilate or contract temporal coordinates end-to-end, our approach captures locally varying periodicities without requiring pre-defined warping functions. We implement this mechanism in StretchTime, a multivariate forecasting architecture that achieves state-of-the-art performance on standard benchmarks, demonstrating superior robustness on datasets exhibiting non-stationary temporal dynamics.Open AccessStretchTime: Adaptive Time Series Forecasting via Symplectic Attention
Transformer architectures have established strong baselines in time series forecasting, yet they typically rely on positional encodings that assume uniform, index-based temporal progression. However, real-world systems, from shifting financial cycles to elastic biological rhythms, frequently exhibit "time-warped" dynamics where the effective flow of time decouples from the sampling index. In this work, we first formalize this misalignment and prove that rotary position embedding (RoPE) is mathematically incapable of representing non-affine temporal warping. To address this, we propose Symplectic Positional Embeddings (SyPE), a learnable encoding framework derived from Hamiltonian mechanics. SyPE strictly generalizes RoPE by extending the rotation group $\mathrm{SO}(2)$ to the symplectic group $\mathrm{Sp}(2,\mathbb{R})$, modulated by a novel input-dependent adaptive warp module. By allowing the attention mechanism to adaptively dilate or contract temporal coordinates end-to-end, our approach captures locally varying periodicities without requiring pre-defined warping functions. We implement this mechanism in StretchTime, a multivariate forecasting architecture that achieves state-of-the-art performance on standard benchmarks, demonstrating superior robustness on datasets exhibiting non-stationary temporal dynamics. - Open Access
Free Energy Mixer
Standard attention stores keys/values losslessly but reads them via a per-head convex average, blocking channel-wise selection. We propose the Free Energy Mixer (FEM): a free-energy (log-sum-exp) read that applies a value-driven, per-channel log-linear tilt to a fast prior (e.g., from queries/keys in standard attention) over indices. Unlike methods that attempt to improve and enrich the $(q,k)$ scoring distribution, FEM treats it as a prior and yields a value-aware posterior read at unchanged complexity, smoothly moving from averaging to per-channel selection as the learnable inverse temperature increases, while still preserving parallelism and the original asymptotic complexity ($O(T^2)$ for softmax; $O(T)$ for linearizable variants). We instantiate a two-level gated FEM that is plug-and-play with standard and linear attention, linear RNNs and SSMs. It consistently outperforms strong baselines on NLP, vision, and time-series at matched parameter budgets.Open AccessFree Energy Mixer
Standard attention stores keys/values losslessly but reads them via a per-head convex average, blocking channel-wise selection. We propose the Free Energy Mixer (FEM): a free-energy (log-sum-exp) read that applies a value-driven, per-channel log-linear tilt to a fast prior (e.g., from queries/keys in standard attention) over indices. Unlike methods that attempt to improve and enrich the $(q,k)$ scoring distribution, FEM treats it as a prior and yields a value-aware posterior read at unchanged complexity, smoothly moving from averaging to per-channel selection as the learnable inverse temperature increases, while still preserving parallelism and the original asymptotic complexity ($O(T^2)$ for softmax; $O(T)$ for linearizable variants). We instantiate a two-level gated FEM that is plug-and-play with standard and linear attention, linear RNNs and SSMs. It consistently outperforms strong baselines on NLP, vision, and time-series at matched parameter budgets. - Open Access
ZeroS: Zero-Sum Linear Attention for Efficient Transformers
Linear attention methods offer Transformers $O(N)$ complexity but typically underperform standard softmax attention. We identify two fundamental limitations affecting these approaches: the restriction to convex combinations that only permits additive information blending, and uniform accumulated weight bias that dilutes attention in long contexts. We propose Zero-Sum Linear Attention (ZeroS), which addresses these limitations by removing the constant zero-order term $1/t$ and reweighting the remaining zero-sum softmax residuals. This modification creates mathematically stable weights, enabling both positive and negative values and allowing a single attention layer to perform contrastive operations. While maintaining $O(N)$ complexity, ZeroS theoretically expands the set of representable functions compared to convex combinations. Empirically, it matches or exceeds standard softmax attention across various sequence modeling benchmarks.Open AccessZeroS: Zero-Sum Linear Attention for Efficient Transformers
Linear attention methods offer Transformers $O(N)$ complexity but typically underperform standard softmax attention. We identify two fundamental limitations affecting these approaches: the restriction to convex combinations that only permits additive information blending, and uniform accumulated weight bias that dilutes attention in long contexts. We propose Zero-Sum Linear Attention (ZeroS), which addresses these limitations by removing the constant zero-order term $1/t$ and reweighting the remaining zero-sum softmax residuals. This modification creates mathematically stable weights, enabling both positive and negative values and allowing a single attention layer to perform contrastive operations. While maintaining $O(N)$ complexity, ZeroS theoretically expands the set of representable functions compared to convex combinations. Empirically, it matches or exceeds standard softmax attention across various sequence modeling benchmarks. - Open Access
CAPS: Unifying Attention, Recurrence, and Alignment in Transformer-based Time Series Forecasting
This paper presents $\textbf{CAPS}$ (Clock-weighted Aggregation with Prefix-products and Softmax), a structured attention mechanism for time series forecasting that decouples three distinct temporal structures: global trends, local shocks, and seasonal patterns. Standard softmax attention entangles these through global normalization, while recent recurrent models sacrifice long-term, order-independent selection for order-dependent causal structure. CAPS combines SO(2) rotations for phase alignment with three additive gating paths -- Riemann softmax, prefix-product gates, and a Clock baseline -- within a single attention layer. We introduce the Clock mechanism, a learned temporal weighting that modulates these paths through a shared notion of temporal importance. Experiments on long- and short-term forecasting benchmarks surpass vanilla softmax and linear attention mechanisms and demonstrate competitive performance against seven strong baselines with linear complexity. Our code implementation is available at https://github.com/vireshpati/CAPS-Attention.Open AccessCAPS: Unifying Attention, Recurrence, and Alignment in Transformer-based Time Series Forecasting
This paper presents $\textbf{CAPS}$ (Clock-weighted Aggregation with Prefix-products and Softmax), a structured attention mechanism for time series forecasting that decouples three distinct temporal structures: global trends, local shocks, and seasonal patterns. Standard softmax attention entangles these through global normalization, while recent recurrent models sacrifice long-term, order-independent selection for order-dependent causal structure. CAPS combines SO(2) rotations for phase alignment with three additive gating paths -- Riemann softmax, prefix-product gates, and a Clock baseline -- within a single attention layer. We introduce the Clock mechanism, a learned temporal weighting that modulates these paths through a shared notion of temporal importance. Experiments on long- and short-term forecasting benchmarks surpass vanilla softmax and linear attention mechanisms and demonstrate competitive performance against seven strong baselines with linear complexity. Our code implementation is available at https://github.com/vireshpati/CAPS-Attention. - Open Access
A Prospective Real-time Early Warning System to Anticipate Onsets and Peaks of Respiratory Diseases Outbreaks at the State Level in the U.S. A Transfer Learning Approach Leveraging Digital Traces
Respiratory disease outbreaks burden U.S. healthcare systems with over one million hospitalizations annually, yet current surveillance systems lag 1-2 weeks behind real-time conditions, preventing timely intervention. We developed a machine learning early warning system that combines Google search trends with traditional epidemiological data using ensemble voting algorithms to predict the timing of outbreak onsets and peaks across multiple respiratory pathogens. The system applies anomaly detection and transfer learning to monitor syndromic Influenza-like illnesses (ILI), and hospitalizations caused by respiratory syncytial virus (RSV) or Influenza, simultaneously, across all 50 US states. During operational real-time deployment from August 2024 through the 2024-2025 season, the system detected 98.0% of outbreak onsets with 5-week average lead time and 97.0% of peaks with 2-week average lead time, achieving positive predictive values that exceed 82%. This framework transforms reactive public health responses into proactive epidemic preparedness by reducing historical timing uncertainty from 10-20 weeks to consistent 2-6 week prediction windows, providing a scalable approach for monitoring both seasonal outbreaks and emerging respiratory threats.
Open AccessA Prospective Real-time Early Warning System to Anticipate Onsets and Peaks of Respiratory Diseases Outbreaks at the State Level in the U.S. A Transfer Learning Approach Leveraging Digital Traces
Respiratory disease outbreaks burden U.S. healthcare systems with over one million hospitalizations annually, yet current surveillance systems lag 1-2 weeks behind real-time conditions, preventing timely intervention. We developed a machine learning early warning system that combines Google search trends with traditional epidemiological data using ensemble voting algorithms to predict the timing of outbreak onsets and peaks across multiple respiratory pathogens. The system applies anomaly detection and transfer learning to monitor syndromic Influenza-like illnesses (ILI), and hospitalizations caused by respiratory syncytial virus (RSV) or Influenza, simultaneously, across all 50 US states. During operational real-time deployment from August 2024 through the 2024-2025 season, the system detected 98.0% of outbreak onsets with 5-week average lead time and 97.0% of peaks with 2-week average lead time, achieving positive predictive values that exceed 82%. This framework transforms reactive public health responses into proactive epidemic preparedness by reducing historical timing uncertainty from 10-20 weeks to consistent 2-6 week prediction windows, providing a scalable approach for monitoring both seasonal outbreaks and emerging respiratory threats.