Summary of United We Pretrain, Divided We Fail! Representation Learning For Time Series by Pretraining on 75 Datasets at Once, By Maurice Kraus and Felix Divo and David Steinmann and Devendra Singh Dhami and Kristian Kersting
United We Pretrain, Divided We Fail! Representation Learning for Time Series by Pretraining on 75 Datasets at Once
by Maurice Kraus, Felix Divo, David Steinmann, Devendra Singh Dhami, Kristian Kersting
First submitted to arxiv on: 23 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 new self-supervised contrastive pretraining approach learns a single encoding from many unlabeled and diverse time series datasets, which can then be reused in several target domains for tasks like classification. The proposed XD-MixUp interpolation method and Soft Interpolation Contextual Contrasting (SICC) loss enable this learning process. Compared to supervised training and other self-supervised pretraining methods, this approach outperforms when finetuning on low-data regimes. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research shows that we can actually learn from many time series datasets, even from 75 at once! A new method is introduced to learn one encoding from multiple unlabeled time series datasets. This learned representation can be used in different target domains for tasks like classification. The approach outperforms other methods when working with limited data. |
Keywords
* Artificial intelligence * Classification * Pretraining * Self supervised * Supervised * Time series
