Retrieval Is Not Enough: Refreshing Memory for Frozen Time-Series Forecasters

Key points

• Retrieval-augmented time-series forecasting uses the continuations of historical segments similar to the current context as references for a forecaster.1
• Most existing methods build the retrieval memory once from the training segment, leaving observations revealed after deployment unavailable as references, and generally do not calibrate how much the retrieved information should influence a frozen forecaster.1
• We identify two key determinants of retrieval utility for a frozen forecaster: whether the history still reflects the current state, and whether the correction it induces aligns with the forecaster's residual errors, an alignment that can shift between validation and deployment when the memory becomes stale.1
• Under a simplified generative model, we characterize the optimal combination gain through the second-order relation between forecaster error and memory correction, and show that a sufficiently long look-back can make periodic memory information redundant.1

Sources (1)

• [1]Retrieval Is Not Enough: Refreshing Memory for Frozen Time-Series Forecasters
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 06:33 AM

  We propose FreshCast, a plug-in retrieval framework that keeps the forecaster frozen, continuously updates a non-parametric memory with new observations, forms a memory forecast through relational kernel regression, and calibrates its weight in closed form on the validation segment.

  Retrieval-augmented time-series forecasting uses the continuations of historical segments similar to the current context as references for a forecaster.

Extractive summary: sentences quoted from the sources.