Overtone: Cyclic Patch Modulation for Clean, Efficient, and Flexible Physics Emulators
Transformer-based PDE surrogates achieve remarkable performance but face two key challenges: fixed patch sizes cause systematic error accumulation at harmonic frequencies, and computational costs remain inflexible regardless of problem complexity or available resources. We introduce Overtone, a unified solution through dynamic patch size control at inference. Overtone's key insight is that cyclically modulating patch sizes during autoregressive rollouts distributes errors across the frequency spectrum, mitigating the systematic harmonic artifact accumulation that plague fixed-patch models. We implement this through two architecture-agnostic modules--CSM (using dynamic stride modulation) and CKM (using dynamic kernel resizing)--that together provide both harmonic mitigation and compute-adaptive deployment. This flexible tokenization lets users trade accuracy for speed dynamically based on computational constraints, and the cyclic rollout strategy yields up to 40% lower long rollout error in variance-normalised RMSE (VRMSE) compared to conventional, static-patch surrogates. Across challenging 2D and 3D PDE benchmarks, one Overtone model matches or exceeds fixed-patch baselines across inference compute budgets, when trained under a fixed total training budget setting.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
StARS DCM: A Sleep Stage-Decoding Forehead EEG Patch for Real-time Modulation of Sleep Physiology
The System to Augment Restorative Sleep (StARS) is a modular hardware/software platform designed for real-time sleep monitoring and intervention. Utilizing the compact DCM biosignal device, StARS captures electrophysiolo…
EEGTransfer LearningA 300-500 MHz Tunable Oscillator Exploiting Ten Overtones in Single Lithium Niobate Resonator
This paper presents the first voltage-controlled MEMS oscillator (VCMO) based on a Lithium Niobate (LiNbO3) lateral overtone bulk acoustic resonator (LOBAR). The VCMO consists of a LOBAR in a closed loop with 2 amplifica…
Machine learning empowered Modulation detection for OFDM-based signals
We propose a blind ML-based modulation detection for OFDM-based technologies. Unlike previous works that assume an ideal environment with precise knowledge of subcarrier count and cyclic prefix location, we consider blin…
PAMod: Modeling Cyclical Shifts via Phase-Amplitude Modulation for Non-stationary Time Series Forecasting
Real-world time series forecasting faces the fundamental challenge of non-stationary statistical properties, including shifts in mean and variance over time. While reversible instance normalization (RevIN) has shown prom…
Time Series ForecastingRepresentation LearningFeature Extraction, Modulation and Recognition of Mixed Signal Based on SVM
This paper introduces likelihood-based and feature-based modulation recognition methods. In the feature-based modulation simulation part, instantaneous feature, cyclic spectrum, high-order cumulants, and wavelet transfor…