Meta Guidance: Incorporating Inductive Biases into Deep Time Series Imputers

Liqiang Nie (Harbin Institute of Technology (Shenzhen)) · Xinyang Chen (Tsinghua University) · Weili Guan (Harbin Institute of Technology (Shenzhen)) · Jiacheng You (Harbin Institute of Technology) · Yu Sun (California Polytechnic State University, Pomona)

benchmark datasetsdata-adaptive weightsdeep imputation modelsimputation errorinductive biaseslearnable weight allocationmeta guidancemissing values non-stationaritynon-stationary guidanceperiodic guidanceperiodicityproximity principlestate-of-the-art baselinestime series data

Missing values, frequently encountered in time series data, can significantly impair the effectiveness of analytical methods. While deep imputation models have emerged as the predominant approach due to their superior performance, explicitly incorporating inductive biases aligned with time-series characteristics offers substantial improvement potential. Taking advantage of non-stationarity and periodicity in time series, two domain-specific inductive biases are designed: (1) Non-Stationary Guidance, which operationalizes the proximity principle to address highly non-stationary series by emphasizing temporal neighbors, and (2) Periodic Guidance, which exploits periodicity patterns through learnable weight allocation across historical periods. Building upon these complementary mechanisms, the overall module, named Meta Guidance, dynamically fuses both guidances through data-adaptive weights learned from the specific input sample. Experiments on nine benchmark datasets demonstrate that integrating Meta Guidance into existing deep imputation architectures achieves an average 27.39\% reduction in imputation error compared to state-of-the-art baselines.