How Mask Matters: Towards Theoretical Understandings of Masked Autoencoders

Authors: Qi Zhang, Yifei Wang, Yisen Wang

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we first present the main empirical results of our proposed U-MAE loss on different real-world datasets with different backbones. Then we conduct a series of experiments to understand how well the U-MAE loss works.
Researcher Affiliation Academia 1 Key Lab. of Machine Perception (Mo E), School of Intelligence Science and Technology, Peking University 2 School of Mathematical Sciences, Peking University 3 Institute for Artificial Intelligence, Peking University
Pseudocode No No structured pseudocode or algorithm blocks were found in the paper.
Open Source Code Yes Code is available at https://github.com/zhangq327/U-MAE.
Open Datasets Yes extensive experiments are conducted on CIFAR-10 [21], Image Net-100 [8], and Image Net-1K [8].
Dataset Splits No No explicit details on train/validation/test splits (e.g., percentages, sample counts, or citations to predefined splits) were found in the main text.
Hardware Specification No No specific hardware details (e.g., exact GPU/CPU models, memory amounts, or detailed computer specifications) were found.
Software Dependencies No No specific ancillary software details, such as library or solver names with version numbers, were found in the paper.
Experiment Setup Yes The mask ratio is set to 0.75. For U-MAE, the coefficient of the uniformity term is set to 0.01. On CIFAR-10, we pretrain the model for 2000 epochs with batch size 4096 and weight decay 0.05. On Image Net-100 and Image Net-1K, we pretrain the model for 200 epochs with batch size 1024 and weight decay 0.05.