Summary of Temporal Test-time Adaptation with State-space Models, by Mona Schirmer et al.
Temporal Test-Time Adaptation with State-Space Models
by Mona Schirmer, Dan Zhang, Eric Nalisnick
First submitted to arxiv on: 17 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML)
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 addresses the issue of performance decay in deployed machine learning models due to inevitable distribution shifts between training and test data over time. Existing test-time adaptation methods have primarily focused on synthetic corruption shifts, leaving gradual temporal distribution shifts underexplored. The authors propose STAD, a probabilistic state-space model that adapts a deployed model to these shifts by learning the time-varying dynamics in hidden features. Without requiring labels, STAD infers time-evolving class prototypes acting as a dynamic classification head. Experimental results on real-world temporal distribution shifts demonstrate the effectiveness of STAD in handling small batch sizes and label shift. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary In this paper, scientists try to solve a problem that happens when machine learning models are used in real life. The models don’t always perform well because the data they see is different from the data they were trained on. They propose a new way to fix this called STAD. It’s like having a special filter that helps the model understand how things change over time, without needing extra information or labels. This is important because it can help small groups of examples and when there are changes in what we call “labels”. The results show that STAD works well. |
Keywords
» Artificial intelligence » Classification » Machine learning
