Poster
in
Workshop: AI for Science: Scaling in AI for Scientific Discovery
Bayesian Optimization for the Discovery of Redox Active Quinones
Giacomo De Gobbi · Reyhan Yagmur · Janine Maier · Stefan Spirk · Robert Peharz
Keywords: [ AI For Science ] [ redox-flow batteries ] [ Molecular Design ] [ Bayesian Optimization ]
Traditional computational chemistry techniques are often a severe bottleneck in scaling up the discovery of materials or new useful molecules. Machine learning techniques have proven effective to overcome such limitations and often lead to unprecedented results. In this work we focus on finding candidate molecules that are benzoquinones derivatives to be used in organic flow batteries. We present a sampling algorithm equipped with chemistry-based constraints in order to generate a molecular library and we utilize a Bayesian optimization strategy to select the best suited molecules.