Summary of Reprune: Channel Pruning Via Kernel Representative Selection, by Mincheol Park et al.
REPrune: Channel Pruning via Kernel Representative Selection
by Mincheol Park, Dongjin Kim, Cheonjun Park, Yuna Park, Gyeong Eun Gong, Won Woo Ro, Suhyun Kim
First submitted to arxiv on: 27 Feb 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- 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 A novel channel pruning technique called REPrune is proposed to accelerate convolutional neural networks (CNNs) while maintaining their accuracy. Unlike traditional pruning methods that operate at the unit of a convolution filter, REPrune exploits finer but structured granularity by identifying similar kernels within each channel using agglomerative clustering. The technique then selects filters that maximize the incorporation of kernel representatives while optimizing the maximum cluster coverage problem. By integrating with simultaneous training-pruning paradigm, REPrune promotes efficient and progressive pruning throughout training CNNs, avoiding the conventional train-prune-finetune sequence. Experimental results show that REPrune outperforms existing methods in computer vision tasks, achieving a balance between acceleration ratio and performance retention. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary REPrune is a new way to make convolutional neural networks (CNNs) run faster without losing accuracy. Normally, pruning works by removing parts of the network one filter at a time. But REPrune looks at groups of filters that do similar things and removes or reduces those groups instead. This makes the process more flexible and helps maintain performance. By combining this new method with training, REPrune can make the network faster throughout the learning process without needing to retrain it from scratch. The results show that REPrune is better than other methods for computer vision tasks, finding a good balance between speed and accuracy. |
Keywords
» Artificial intelligence » Clustering » Pruning
