FLAME: Flow Enhanced Legendre Memory Models for General Time Series Forecasting

TL;DR

FLAME introduces a lightweight, versatile time series foundation model utilizing Legendre Memory and Normalization Flow to achieve state-of-the-art deterministic and probabilistic forecasting performance.

Contribution

The paper presents FLAME, a novel framework combining Legendre Memory variants and Normalization Flow for efficient, robust, and accurate time series forecasting.

Findings

Achieves state-of-the-art zero-shot performance on benchmarks.

Effectively captures data inductive bias for long-range inference.

Supports both deterministic and probabilistic forecasting.

Abstract

In this work, we introduce FLAME, a family of extremely lightweight and capable Time Series Foundation Models, which support both deterministic and probabilistic forecasting via generative probabilistic modeling, thus ensuring both efficiency and robustness. FLAME utilizes the Legendre Memory for strong generalization capabilities. Through adapting variants of Legendre Memory, i.e., translated Legendre (LegT) and scaled Legendre (LegS), in the Encoding and Decoding phases, FLAME can effectively capture the inherent inductive bias within data and make efficient long-range inferences. To enhance the accuracy of probabilistic forecasting while keeping efficient, FLAME adopts a Normalization Flow based forecasting head, which can model the arbitrarily intricate distributions over the forecasting horizon in a generative manner. Comprehensive experiments on well-recognized benchmarks, including TSFM-Bench and ProbTS, demonstrate the consistent state-of-the-art zero-shot performance of FLAME on both deterministic and probabilistic forecasting tasks.