Summary of Timesiam: a Pre-training Framework For Siamese Time-series Modeling, by Jiaxiang Dong et al.
TimeSiam: A Pre-Training Framework for Siamese Time-Series Modeling
by Jiaxiang Dong, Haixu Wu, Yuxuan Wang, Yunzhong Qiu, Li Zhang, Jianmin Wang, Mingsheng Long
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 This paper proposes a novel self-supervised pre-training framework called TimeSiam for time series data. Building on the idea of Siamese networks, TimeSiam aims to capture intrinsic temporal correlations between past and current subseries by pre-training encoders to reconstruct these subseries. The approach leverages a simple data augmentation method, such as masking, to generate diverse augmented subseries and learn internal time-dependent representations. Additionally, learnable lineage embeddings are introduced to distinguish temporal distance between sampled series, fostering the learning of diverse temporal correlations. TimeSiam outperforms advanced pre-training baselines across 13 standard benchmarks in both intra- and cross-domain scenarios. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary TimeSiam is a new way to train machines to understand time series data without needing labeled examples. It uses a special kind of neural network called a Siamese network, which helps the machine learn about patterns in the data over time. The approach is simple but effective, and it works well on many different types of data. TimeSiam has the potential to make predictions and classify data more accurately than current methods. |
Keywords
* Artificial intelligence * Data augmentation * Neural network * Self supervised * Siamese network * Time series
