Recently a novel framework has been proposed for designing the molecular structure of chemical compounds using both artificial neural networks (ANNs) and mixed integer linear programming (MILP). In the framework, we first define a feature vector of a chemical graph and construct an ANN that maps to a predicted value of a chemical property to . After this, we formulate an MILP that simulates the computation process of from and that of from . Given a target value of the chemical property , we infer a chemical graph such that by solving the MILP. In this paper, we use linear regression to construct a prediction function instead of ANNs. For this, we derive an MILP formulation that simulates the computation process of a prediction function by linear regression. The results of computational experiments suggest our method can infer chemical graphs with around up to 50 non-hydrogen atoms.
View on arXiv