Summary of Is Oracle Pruning the True Oracle?, by Sicheng Feng et al.
Is Oracle Pruning the True Oracle?
by Sicheng Feng, Keda Tao, Huan Wang
First submitted to arxiv on: 28 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
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 investigates the validity of Oracle pruning, a foundation of most neural network pruning methods, on modern deep models through empirical correlation analyses. The authors analyze the model performance correlation before and after retraining across various models (LeNet5, VGG, ResNets, ViT, MLLM) and datasets (MNIST, CIFAR10/CIFAR100, ImageNet-1K, MLLM data). The results show that the performance before retraining is barely correlated with the performance after retraining, suggesting that Oracle pruning may not guarantee good performance. This finding implies that existing works using Oracle pruning to derive pruning criteria might be groundless. Furthermore, the authors argue that the rising task complexity is one factor making Oracle pruning invalid. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how well Oracle pruning works on modern deep learning models. It seems that Oracle pruning doesn’t really predict how well a model will perform after retraining. This means that some methods for pruning neural networks might not be as good as people thought. The authors also think that the reason why Oracle pruning isn’t working is because of the increasing complexity of tasks. |
Keywords
» Artificial intelligence » Deep learning » Neural network » Pruning » Vit
