Summary of Always-sparse Training by Growing Connections with Guided Stochastic Exploration, By Mike Heddes et al.
Always-Sparse Training by Growing Connections with Guided Stochastic Exploration
by Mike Heddes, Narayan Srinivasa, Tony Givargis, Alexandru Nicolau
First submitted to arxiv on: 12 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 an efficient algorithm for training artificial neural networks (ANNs) that are always sparse, unlike existing methods that only apply sparsity during inference. The new approach offers linear time complexity with respect to the model width during both training and inference, making it scalable to larger models. Moreover, the algorithm improves accuracy over previous sparse training methods through guided stochastic exploration. The proposed method is evaluated on various benchmark datasets using popular architectures like ResNet, VGG, and ViT. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper develops a new way of training artificial neural networks that saves time and energy while keeping the same level of accuracy. Currently, most models are fully trained and then made more efficient by removing some parts, but this approach is not very effective. The researchers introduce an algorithm that trains models to be sparse from the start, which makes them much faster and more efficient. This new method also improves how well the models work compared to previous attempts at making models more efficient during training. |
Keywords
* Artificial intelligence * Inference * Resnet * Vit
