(open) 12-months postdoc to start before September 2021
topic: Machine-learning-assisted discovery of photoresponsive materials for efficient gas release
The postdoc is funded by the MIAI. This project aims at developing an efficient ML strategy to accurately predict carbon capture at low pressure, something that represents a true challenge today owing to the complex adsorption mechanism. Due to the vast and diverse set of topologies, and possibly adsorption mechanisms in MOFs, we will develop a clustering approach based on cheap descriptors in order to optimize the prediction. This will allow to build the best training set but at a limited cost. On this small learning set a full ML method will be applied using variable selection tools, or causality inference, to provide the best regression model to predict CO2 capture.