Fast ES-RNN: A GPU Implementation of the ES-RNN Algorithm

Andrew Redd, Kaung Khin, Aldo Marini
7/7/2019
cs.LGstat.ML

Official code: kgmonkhin/ESRNN-GPU (317 stars, MIT, PyTorch, last commit 2019-09-02).

Abstract

Due to their prevalence, time series forecasting is crucial in multiple domains. We seek to make state-of-the-art forecasting fast, accessible, and generalizable. ES-RNN is a hybrid between classical state space forecasting models and modern RNNs that achieved a 9.4% sMAPE improvement in the M4 competition. Crucially, ES-RNN implementation requires per-time series parameters. By vectorizing the original implementation and porting the algorithm to a GPU, we achieve up to 322x training speedup depending on batch size with similar results as those reported in the original submission. Our code can be found at: https://github.com/damitkwr/ESRNN-GPU

View on arXivView PDF

Code Implementations(1)

kgmonkhin/ESRNN-GPUOfficial100%
PyTorchMIT

PyTorch GPU implementation of the ES-RNN model for time series forecasting

31773R, PythonNov 6, 20187 years agoMIT
deep-forecastingdeep-learninges-rnnforecastingpytorch+1 more

Similar Papers

SST: Multi-Scale Hybrid Mamba-Transformer Experts for Time Series Forecasting

87% match1 repo2342024

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

87% match1 repo12024

Unlocking the Value of Text: Event-Driven Reasoning and Multi-Level Alignment for Time Series Forecasting

87% match1 repo222026

Enhancing Algorithm Performance Understanding through tsMorph: Generating Semi-Synthetic Time Series for Robust Forecasting Evaluation

87% match1 repo2023

BALM-TSF: Balanced Multimodal Alignment for LLM-Based Time Series Forecasting

86% match1 repo172025

LLM4TS: Aligning Pre-Trained LLMs as Data-Efficient Time-Series Forecasters

86% match1 repo342023

Cite this paper

@article{redd2019fast,
  title  = {Fast ES-RNN: A GPU Implementation of the ES-RNN Algorithm},
  author = {Andrew Redd and Kaung Khin and Aldo Marini},
  year   = {2019},
  eprint = {1907.03329},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1907.03329}
}

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/1907.03329.svg)](https://codeofpaper.com/paper/1907.03329)

Discussion