Summary of Learning Symbolic Model-agnostic Loss Functions Via Meta-learning, by Christian Raymond et al.
Learning Symbolic Model-Agnostic Loss Functions via Meta-Learning
by Christian Raymond, Qi Chen, Bing Xue, Mengjie Zhang
First submitted to arxiv on: 19 Sep 2022
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 The paper proposes a new meta-learning framework for learning model-agnostic loss functions, which can significantly improve the performance of models trained under them. The framework combines evolution-based methods with neuro-symbolic search to find symbolic loss functions that are then optimized using end-to-end gradient-based training. Empirically, the proposed method outperforms cross-entropy loss and state-of-the-art loss function learning on various neural network architectures and datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about a new way to learn what makes models work well or poorly. It’s like finding a recipe that makes your favorite cake taste better. The authors developed a special way to search for this “recipe” using both math and computer programming. They tested it on many different types of models and datasets, and it worked really well! This could help make computers even smarter in the future. |
Keywords
* Artificial intelligence * Cross entropy * Loss function * Meta learning * Neural network
