Summary of Blending Low and High-level Semantics Of Time Series For Better Masked Time Series Generation, by Johan Vik Mathisen et al.
Blending Low and High-Level Semantics of Time Series for Better Masked Time Series Generation
by Johan Vik Mathisen, Erlend Lokna, Daesoo Lee, Erlend Aune
First submitted to arxiv on: 29 Aug 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 State-of-the-art time series generation (TSG) approaches, such as TimeVQVAE, employ vector quantization-based tokenization to model complex distributions. These methods transform time series into discrete latent vectors and learn a prior model on the sequence. However, these vectors only capture low-level semantics like shapes. We propose NC-VQVAE, a novel framework that integrates self-supervised learning to derive a discrete latent space capturing both low and high-level semantics. Our approach demonstrates a significant improvement in synthetic sample quality. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Researchers have developed ways to generate fake time series data that looks real. These methods are good at creating simple patterns, but not complex ones. We think we can do better by combining two ideas: learning from the original data and using more detailed information. Our new approach, NC-VQVAE, does just that and makes much better synthetic samples. |
Keywords
» Artificial intelligence » Latent space » Quantization » Self supervised » Semantics » Time series » Tokenization
