Summary of Rose: Register Assisted General Time Series Forecasting with Decomposed Frequency Learning, by Yihang Wang et al.
ROSE: Register Assisted General Time Series Forecasting with Decomposed Frequency Learning
by Yihang Wang, Yuying Qiu, Peng Chen, Kai Zhao, Yang Shu, Zhongwen Rao, Lujia Pan, Bin Yang, Chenjuan Guo
First submitted to arxiv on: 24 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Machine Learning (stat.ML)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed Register Assisted General Time Series Forecasting Model with Decomposed Frequency Learning (ROSE) is a novel pre-trained model for time series forecasting that enables general time series forecasting by obtaining unified representations from multi-domain time-series datasets and capturing domain-specific features. ROSE employs Decomposed Frequency Learning for the pre-training task, which decomposes coupled semantic and periodic information in time series with frequency-based masking and reconstruction to obtain unified representations across domains. Additionally, it learns a Time Series Register that generates a register codebook to capture domain-specific representations during pre-training and enhances domain-adaptive transfer by selecting related register tokens on downstream tasks. ROSE achieves state-of-the-art forecasting performance on 8 real-world benchmarks after pre-training on large-scale time series data. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary ROSE is a new way of predicting the future based on patterns in past data. It takes in lots of different kinds of data and uses that to make good predictions for other types of data. This helps solve two big problems: how to understand what’s similar between different types of data, and how to learn special skills that help with each type of data. ROSE does this by using a clever technique called Decomposed Frequency Learning that looks at the patterns in the data and separates out what’s common from what’s unique. It also learns a special kind of codebook that helps it remember important details about each type of data. After learning all this, ROSE is really good at making predictions – better than many other methods. |
Keywords
* Artificial intelligence * Time series
