Dielectric Polymer Property Prediction Using Recurrent Neural Networks with Optimizations

Despite the growing success of machine learning for predicting structure-property relationships in molecules and materials, such as predicting the dielectric properties of polymers, it is still in its infancy. We report on the effectiveness of solving structure-property relationships for a computer-generated database of dielectric polymers using recurrent neural network (RNN) models. The implementation of a series of optimization strategies was crucial to achieving high learning speeds and sufficient accuracy: (1) binary and nonbinary representations of SMILES (Simplified Molecular Input Line System) fingerprints and (2) backpropagation with affine transformation of the input sequence (ATransformedBP) and resilient backpropagation with initial weight update parameter optimizations (iRPROP(-) optimized). For the investigated database of polymers, the binary SMILES representation was found to be superior to the decimal representation with respect to the training and prediction performance. All developed and optimized Elman-type RNN algorithms outperformed nonoptimized RNN models in the efficient prediction of nonlinear structure-activity relationships. The average relative standard deviation (RSD) remained well below 5%, and the maximum RSD did not exceed 30%. Moreover, we provide a C++ codebase as a testbed for a new generation of open programming languages that target increasingly diverse computer architectures.

Automated summary

The study investigates the effectiveness of using recurrent neural network models to solve structure-property relationships in dielectric polymers, highlighting the importance of optimization strategies for achieving high learning speeds and accuracy. It was found that binary SMILES representation outperformed decimal representation and the developed Elman-type RNN algorithms showed superior performance in predicting nonlinear structure-activity relationships, with average relative standard deviation remaining below 5% and maximum RSD not exceeding 30%. A C++ codebase is also provided as a testbed for new programming languages targeting diverse computer architectures.