The Polymer Society of Korea

Abstract

The intersection of artificial intelligence and polymer science represents one of the most promising frontiers in materials research, yet the complexity of polymer structures has historically posed significant challenges for computational approaches. Traditional molecular representations often struggle to capture the hierarchical nature and intricate connectivity patterns that define polymer behavior, limiting the effectiveness of machine learning models in understanding structure-property relationships. We introduce HAPPY (Hierarchically Abstracted rePeat unit of PolYmers), a revolutionary paradigm that enables Transformer-based AI systems to comprehend the context of polymer chemistry. Our research explores how this paradigm shift enables both forward property prediction and inverse design challenges, opening new possibilities for accelerated polymer discovery.