Summary of Task-customized Masked Autoencoder Via Mixture Of Cluster-conditional Experts, by Zhili Liu et al.
Task-customized Masked AutoEncoder via Mixture of Cluster-conditional Experts
by Zhili Liu, Kai Chen, Jianhua Han, Lanqing Hong, Hang Xu, Zhenguo Li, James T. Kwok
First submitted to arxiv on: 8 Feb 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 Mixture of Cluster-conditional Experts (MoCE) paradigm improves upon the Masked Autoencoder~(MAE) method for pre-training models, addressing the issue of negative transfer when applying pre-trained models to downstream tasks with different data distributions. MoCE trains each expert model on semantically relevant images, allocating customized models to specific downstream tasks. This approach outperforms vanilla MAE by 2.45% on average and achieves new state-of-the-art results in detection and segmentation. |
Low | GrooveSquid.com (original content) | Low Difficulty Summary MoCE is a way to make pre-trained models better for different tasks. Right now, pre-training methods like Masked Autoencoder~(MAE) can be useful but might not work well if the task has different data. To fix this, MoCE trains each expert model only on images that are relevant to the task. This way, each task gets its own customized model pre-trained with similar data. Experiments show that MoCE is better than MAE and sets new records in some areas. |
Keywords
* Artificial intelligence * Autoencoder * Mae