Summary of Less Is More: Pseudo-label Filtering For Continual Test-time Adaptation, by Jiayao Tan et al.
Less is More: Pseudo-Label Filtering for Continual Test-Time Adaptation
by Jiayao Tan, Fan Lyu, Chenggong Ni, Tingliang Feng, Fuyuan Hu, Zhang Zhang, Shaochuang Zhao, Liang Wang
First submitted to arxiv on: 3 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 paper proposes Continual Test-Time Adaptation (CTTA), a method that adapts a pre-trained model to new, unseen domains during the test phase without access to source data. The approach relies on constructing pseudo-labels for unlabeled data and updating the model through self-training. However, existing methods often produce noisy pseudo-labels, leading to insufficient adaptation. To address this issue, the authors propose a Pseudo Labeling Filter (PLF) that selects reliable pseudo-labels for self-training. PLF uses three principles – initialization, growth, and diversity – to set thresholds for filtering pseudo-labels. Additionally, the paper introduces Class Prior Alignment (CPA), which encourages the model to make diverse predictions for unknown domain samples. The authors demonstrate the effectiveness of PLF through extensive experiments, outperforming current state-of-the-art methods in CTTA. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about a new way to help artificial intelligence models learn from new data during testing without needing the original training data. It’s called Continual Test-Time Adaptation (CTTA). Right now, models are good at learning from data they’ve seen before, but struggle when faced with new, unknown types of data. To fix this problem, the authors developed a method to create “fake” labels for new data and use them to update the model. But these fake labels can be noisy, which makes it hard for the model to learn. The authors came up with a way to filter out the bad fake labels and only use the good ones. This helps the model adapt better to new types of data. |
Keywords
» Artificial intelligence » Alignment » Self training
