Summary of Unsupervised Concept Drift Detection From Deep Learning Representations in Real-time, by Salvatore Greco et al.
Unsupervised Concept Drift Detection from Deep Learning Representations in Real-time
by Salvatore Greco, Bartolomeo Vacchetti, Daniele Apiletti, Tania Cerquitelli
First submitted to arxiv on: 24 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 proposed DriftLens framework is an unsupervised real-time concept drift detection system that can effectively detect and characterize concept drift in unstructured data using deep learning representations. It outperforms previous methods in detecting drift in 11 out of 13 use cases, runs at least 5 times faster, and provides coherent results with the amount of drift (correlation ≥ 0.85). The framework can be applied to various domains such as text, image, and speech classification tasks. By leveraging distribution distances of deep learning representations, DriftLens addresses the limitations of previous methods, which often require ground-truth labels or are complex and difficult to implement in real-world scenarios. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary DriftLens is a new way to detect when a model’s performance changes over time. This happens because the data it’s trained on changes too. Most current methods need labels to work, but these aren’t always available. DriftLens is different because it can detect drift without those labels. It does this by looking at how similar or different the representations of deep learning are. The results show that DriftLens works better than other methods in 11 out of 13 cases and runs faster too. |
Keywords
* Artificial intelligence * Classification * Deep learning * Unsupervised
