Summary of In-context Time Series Predictor, by Jiecheng Lu et al.
In-context Time Series Predictor
by Jiecheng Lu, Yan Sun, Shihao Yang
First submitted to arxiv on: 23 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); 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 This paper presents a novel approach for time series forecasting (TSF) using large language models (LLMs). Unlike previous methods, the proposed method formulates TSF tasks as input tokens by constructing pairs of lookback and future values. This reformulation aligns with the in-context learning mechanism of LLMs, making it more parameter-efficient and addressing overfitting issues in existing Transformer-based TSF models. The approach is evaluated across full-data, few-shot, and zero-shot settings, achieving better performance compared to previous architectures. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper uses a new way to predict future values based on past data using really big language models. Instead of treating the prediction problem like a normal machine learning task, it breaks down the time series data into smaller chunks that can be easily processed by the language model. This approach is better than previous methods because it doesn’t require as much training data and is less likely to get stuck in patterns in the past data. |
Keywords
» Artificial intelligence » Few shot » Language model » Machine learning » Overfitting » Parameter efficient » Time series » Transformer » Zero shot
