Summary of Grassmannian Geometry Meets Dynamic Mode Decomposition in Dmd-gen: a New Metric For Mode Collapse in Time Series Generative Models, by Amime Mohamed Aboussalah and Yassine Abbahaddou
Grassmannian Geometry Meets Dynamic Mode Decomposition in DMD-GEN: A New Metric for Mode Collapse in Time Series Generative Models
by Amime Mohamed Aboussalah, Yassine Abbahaddou
First submitted to arxiv on: 15 Dec 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 metric called DMD-GEN to quantify mode collapse in generative models for time series data. Mode collapse refers to the inability of these models to capture the full diversity of their training data, leading to reduced creativity and novelty in generated outputs. The authors introduce a new definition of mode collapse specific to time series data and develop a metric that utilizes Dynamic Mode Decomposition (DMD) to identify coherent spatiotemporal patterns and Optimal Transport to assess discrepancies between original and generated data. This approach not only quantifies the preservation of essential dynamic characteristics but also provides interpretability by pinpointing which modes have collapsed. The authors validate DMD-GEN on both synthetic and real-world datasets using various generative models, including TimeGAN, TimeVAE, and DiffusionTS. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about finding a way to measure how well computer programs can generate new time series data that looks like the real thing. These programs are called generative models, and they’re used in many areas like finance and healthcare. The problem is that these models often don’t do a great job of capturing all the different patterns and trends in the original data. This is called mode collapse, and it means the generated data doesn’t look very realistic. The authors came up with a new way to measure this mode collapse using something called Dynamic Mode Decomposition (DMD). They tested their method on some example datasets and showed that it works well. |
Keywords
» Artificial intelligence » Spatiotemporal » Time series
