Summary of Transferring Core Knowledge Via Learngenes, by Fu Feng et al.
Transferring Core Knowledge via Learngenes
by Fu Feng, Jing Wang, Xin Geng
First submitted to arxiv on: 16 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Neural and Evolutionary Computing (cs.NE)
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 proposes Genetic Transfer Learning (GTL), a framework inspired by the evolutionary process of organisms. GTL trains a population of neural networks, selects superior “learngenes” through tournaments, applies mutations to these genes, and passes them to next generations. By leveraging this concept, the authors demonstrate that learngenes can transfer core knowledge between networks, leading to improved performance on downstream tasks with reduced parameter requirements. Specifically, they show that learngenes of VGG11 and ResNet12 can bring 12% and 16% accuracy improvements on CIFAR-FS and miniImageNet, respectively. The authors also highlight the scalability and adaptability of GTL in different network and dataset configurations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary In a nutshell, this paper develops a new way to teach neural networks using a concept called “learngenes.” These genes are like instructions that help a network learn better and faster. The researchers show that by using these genes, they can train networks with fewer parameters, which makes them more efficient. They tested this idea on two types of networks (VGG11 and ResNet12) and found that it worked well. |
Keywords
* Artificial intelligence * Transfer learning
