Summary of Minusformer: Improving Time Series Forecasting by Progressively Learning Residuals, By Daojun Liang et al.
Minusformer: Improving Time Series Forecasting by Progressively Learning Residuals
by Daojun Liang, Haixia Zhang, Dongfeng Yuan, Bingzheng Zhang, Minggao Zhang
First submitted to arxiv on: 4 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 Minusformer model tackles the issue of overfitting in ubiquitous time series forecasting by adopting a de-redundancy approach. It combines a deep Boosting ensemble learning method with a vanilla Transformer, reorienting its information aggregation mechanism from addition to subtraction. This allows for the incorporation of an auxiliary output branch, which facilitates layer-by-layer learning of residuals and implicit progressive decomposition of input and output streams. The result is a model that offers enhanced versatility, interpretability, and resilience against overfitting. Experimental results demonstrate an average performance improvement of 11.9% across various datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The Minusformer model helps solve the problem of time series forecasting models getting too good at fitting the training data and not doing well on new data. It does this by changing how it combines information from different parts of the model. This lets the model learn in a way that’s more like humans do, by focusing on what’s left out after previous steps. The results show that Minusformer is better than other methods at predicting future values. |
Keywords
* Artificial intelligence * Boosting * Overfitting * Time series * Transformer
