Summary of Onenet: Enhancing Time Series Forecasting Models Under Concept Drift by Online Ensembling, By Yi-fan Zhang et al.
OneNet: Enhancing Time Series Forecasting Models under Concept Drift by Online Ensembling
by Yi-Fan Zhang, Qingsong Wen, Xue Wang, Weiqi Chen, Liang Sun, Zhang Zhang, Liang Wang, Rong Jin, Tieniu Tan
First submitted to arxiv on: 22 Sep 2023
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Data Structures and Algorithms (cs.DS)
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 This paper proposes OneNet, an online time series forecasting model that efficiently updates and combines two models to address the concept drifting problem. The first model focuses on modeling dependencies across the time dimension, while the second model exploits cross-variable dependencies. OneNet uses a reinforcement learning-based approach within an online convex programming framework, allowing for dynamically adjusted weights in combining the two models. This addresses the slow adaptation of classical online learning methods to concept drift. Empirical results show that OneNet reduces forecasting error by over 50% compared to the state-of-the-art method. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary OneNet is a new way to update time series forecasts as new data comes in. Currently, many algorithms are used for this task, but they all have their own strengths and weaknesses. OneNet combines two different approaches to make the best use of both. It uses reinforcement learning to adjust how much each approach contributes to the final forecast. This helps the model adapt quickly to changes in the data, which is important because real-world data often changes over time. The results show that OneNet can reduce errors by more than half compared to other methods. |
Keywords
* Artificial intelligence * Online learning * Reinforcement learning * Time series
