Optimization of large reaction systems is crucial for chemical research and industrial production, but represents a significant challenge given the heavy experimental load required to find optimal high-yielding conditions. Here, the authors introduce an efficient machine learning tool, RS-Coreset, where they take advantage of deep representation learning techniques to guide an interactive procedure for representing the full reaction space, enabling yield prediction with small-scale data.
- Peng-Xiang Hua
- Zhen Huang
- Hu Ding