• Funding : Artois
  • Start year :
  • 2020

Machine learning techniques are already used to determine the association constants between a cyclodextrin (CD) and a guest. However, the case of a mixture of CDs is not addressed. We propose in this thesis to use another artificial intelligence approach to predict the association constants of a mixture of CDs and a guest: constraint programming. The size and position of a CD’s clusters and the size and shape of the guest will be variables, defining a space of possible mixtures. The constraints will express knowledge of molecular modeling. Constraint programming can also be used in a second step to ’explain’ the predictions.