Advancing physiological time series reconstruction and imputation via mixture of receptive fields and experts fusion

2025-11-27

AI Summary

Physiological time series—characterized by high noise, multivariate dynamics, strong nonstationarity, and susceptibility to artifacts—pose significant challenges for existing deep learning-based imputation and reconstruction methods, which suffer from limited robustness and computational efficiency. To address this, we propose an efficient and robust framework grounded in score-matching diffusion models. Our key contributions are: (1) the first adaptive-receptive-field mixture-of-experts (RFAMoE) noise estimator, dynamically tailored to local signal nonstationarity; and (2) a fused MoE module that generates K parallel noise estimates in a single forward pass and routes them via learned weighted fusion—eliminating inefficient multi-inference averaging. Evaluated across multiple physiological datasets, our method surpasses state-of-the-art diffusion models in reconstruction accuracy while reducing inference latency by 42% and FLOPs by 38%.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Intelligent Robots: State Estimation
• Computer Vision: Diffusion Models for Vision

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Recent studies show that using diffusion models for time series signal reconstruc- tion holds great promise. However, such approaches remain largely unexplored in the domain of medical time series. The unique characteristics of the physiological time series signals, such as multivariate, high temporal variability, highly noisy, and artifact-prone, make deep learning-based approaches still challenging for tasks such as imputation. Hence, we propose a novel Mixture of Experts (MoE)-based noise estimator within a score-based diffusion framework. Specifically, the Receptive Field Adaptive MoE (RFAMoE) module is designed to enable each channel to adap- tively select desired receptive fields throughout the diffusion process. Moreover, recent literature has found that when generating a physiological signal, performing multiple inferences and averaging the reconstructed signals can effectively reduce reconstruction errors, but at the cost of significant computational and latency over- head. We design a Fusion MoE module and innovatively leverage the nature of MoE module to generate K noise signals in parallel, fuse them using a routing mechanism, and complete signal reconstruction in a single inference step. This design not only improves performance over previous methods but also eliminates the substantial computational cost and latency associated with multiple inference processes. Extensive results demonstrate that our proposed framework consistently outperforms diffusion-based SOTA works on different tasks and datasets.

Problem

• Reconstructs physiological time series with high variability and noise
• Imputes missing data in multivariate medical signals using diffusion models
• Reduces computational cost of signal reconstruction via parallel MoE fusion

Innovation

• Mixture of Experts noise estimator in diffusion framework
• Receptive Field Adaptive MoE for channel-specific receptive fields
• Fusion MoE for parallel noise generation in single inference