Mamba4Cast: Efficient Zero-Shot Time Series Forecasting with State Space Models

Sathya Kamesh Bhethanabhotla, Omar Swelam, Julien Siems, David Salinas, Frank Hutter
10/12/2024
cs.LGcs.AI

Official code: automl/Mamba4Cast (56 stars, MIT, last commit 2024-10-11).

Abstract

This paper introduces Mamba4Cast, a zero-shot foundation model for time series forecasting. Based on the Mamba architecture and inspired by Prior-data Fitted Networks (PFNs), Mamba4Cast generalizes robustly across diverse time series tasks without the need for dataset specific fine-tuning. Mamba4Cast's key innovation lies in its ability to achieve strong zero-shot performance on real-world datasets while having much lower inference times than time series foundation models based on the transformer architecture. Trained solely on synthetic data, the model generates forecasts for entire horizons in a single pass, outpacing traditional auto-regressive approaches. Our experiments show that Mamba4Cast performs competitively against other state-of-the-art foundation models in various data sets while scaling significantly better with the prediction length. The source code can be accessed at https://github.com/automl/Mamba4Cast.

View on arXivView PDF

Code Implementations(1)

automl/Mamba4CastOfficial100%
MIT

Mamba4Cast, a zero-shot time series forecasting model, achieves competitive performance and faster inference than transformer-based models by leveraging Mamba architecture and synthetic training data to generalize robustly without fine-tuning.

563Shell, PythonOct 11, 20242 years agoMIT
linearrnnmambastate-space-modeltime-seriestime-series-forecasting

Similar Papers

Reverso: Efficient Time Series Foundation Models for Zero-shot Forecasting

88% match2026

Only the Curve Shape Matters: Training Foundation Models for Zero-Shot Multivariate Time Series Forecasting through Next Curve Shape Prediction

87% match2024

FinCast: A Foundation Model for Financial Time-Series Forecasting

86% match2025

Time-FFM: Towards LM-Empowered Federated Foundation Model for Time Series Forecasting

86% match1 repo12026

Impermanent: A Live Benchmark for Temporal Generalization in Time Series Forecasting

86% match1 repo132026

VisionTS: Visual Masked Autoencoders Are Free-Lunch Zero-Shot Time Series Forecasters

85% match1 repo2822026

Cite this paper

@article{bhethanabhotla2024mambacast,
  title  = {Mamba4Cast: Efficient Zero-Shot Time Series Forecasting with State Space Models},
  author = {Sathya Kamesh Bhethanabhotla and Omar Swelam and Julien Siems and David Salinas and Frank Hutter},
  year   = {2024},
  eprint = {2410.09385},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2410.09385v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2410.09385.svg)](https://codeofpaper.com/paper/2410.09385)

Discussion