Search papers, labs, and topics across Lattice.
This paper introduces CRAFT, a channel-wise retrieval-augmented forecasting framework for multivariate time series that addresses the limitations of fixed lookback windows and channel-agnostic retrieval. CRAFT uses a two-stage retrieval pipeline: first, a sparse relation graph prunes irrelevant candidates in the time domain, and then spectral similarity in the frequency domain ranks references. Experiments on seven benchmarks show CRAFT outperforms state-of-the-art baselines in accuracy and inference efficiency.
Channel-specific historical data retrieval boosts multivariate time series forecasting accuracy, finally accounting for the fact that different variables have different rhythms.
Multivariate time series forecasting often struggles to capture long-range dependencies due to fixed lookback windows. Retrieval-augmented forecasting addresses this by retrieving historical segments from memory, but existing approaches rely on a channel-agnostic strategy that applies the same references to all variables. This neglects inter-variable heterogeneity, where different channels exhibit distinct periodicities and spectral profiles. We propose CRAFT (Channel-wise retrieval-augmented forecasting), a novel framework that performs retrieval independently for each channel. To ensure efficiency, CRAFT adopts a two-stage pipeline: a sparse relation graph constructed in the time domain prunes irrelevant candidates, and spectral similarity in the frequency domain ranks references, emphasizing dominant periodic components while suppressing noise. Experiments on seven public benchmarks demonstrate that CRAFT outperforms state-of-the-art forecasting baselines, achieving superior accuracy with practical inference efficiency.
