Summary of Guided Evolution with Binary Discriminators For Ml Program Search, by John D. Co-reyes et al.
Guided Evolution with Binary Discriminators for ML Program Search
by John D. Co-Reyes, Yingjie Miao, George Tucker, Aleksandra Faust, Esteban Real
First submitted to arxiv on: 8 Feb 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 a novel approach to automatically design better machine learning (ML) programs, specifically addressing the challenge of using learning itself to guide the search process. The authors introduce a binary discriminator trained online to distinguish between two ML programs, allowing it to select better programs without performing costly evaluations. This speedup is achieved by combining the discriminator with modern Graph Neural Networks (GNNs) and an adaptive mutation strategy. The proposed method can encode various ML components, including symbolic optimizers, neural architectures, reinforcement learning loss functions, and symbolic regression equations, all represented using a directed acyclic graph (DAG). Experiments demonstrate that this approach can accelerate evolution across diverse problems, achieving a 3.7x speedup for symbolic search in ML optimizers and a 4x speedup for RL loss functions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper aims to improve the process of automatically designing better machine learning programs by using learning itself to guide the search. It proposes a new method that uses an online-trained binary discriminator to select better programs without performing costly evaluations, which can speed up the optimization process. The authors also demonstrate how their approach can be applied to various ML components and achieve significant speedups on diverse problems. |
Keywords
* Artificial intelligence * Machine learning * Optimization * Regression * Reinforcement learning
