Summary of Dense Optimizer : An Information Entropy-guided Structural Search Method For Dense-like Neural Network Design, by Liu Tianyuan et al.
Dense Optimizer : An Information Entropy-Guided Structural Search Method for Dense-like Neural Network Design
by Liu Tianyuan, Hou Libin, Wang Linyuan, Song Xiyu, Yan Bin
First submitted to arxiv on: 10 Oct 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); 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 This paper presents an architecture search method called Dense Optimizer for designing efficient and compact dense convolutional networks (DenseNets). The Dense-like architectures are typically designed manually, but this approach becomes challenging as the network size increases. To address this issue, we propose a novel optimization method that views the dense network as a hierarchical information system. Our proposed algorithm integrates a power-law principle with search space scaling to efficiently solve an optimization problem. This approach is validated on various computer vision benchmark datasets and outperforms existing methods in terms of both accuracy and efficiency. Specifically, our searched model, DenseNet-OPT, achieves a top 1 accuracy of 84.3% on CIFAR-100, which is 5.97% higher than the original DenseNet. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper talks about making computer networks better at recognizing images. Normally, these networks are designed by experts, but that can be time-consuming and difficult to do well. The authors propose a new way to design these networks using an algorithm that finds the best combination of building blocks. This method is tested on various image recognition tasks and performs significantly better than traditional methods. For example, it achieves an accuracy of 84.3% on one particular task, which is 5.97% better than the original method. |
Keywords
* Artificial intelligence * Optimization
