| 研究実績の概要 |
By combining Explainable Artificial Intelligence (XAI) principles and self-learning entropic population annealing (SLEPA) method, we can efficiently explore global distribution while ensuring outputs are explainable and transparent. In detail, we first validated the effectiveness of SLEPA by comparing the 10-layer graphene-WS2 heterostructure’s thermal conductivity distribution among ground truth, SLEPA, Bayesian optimization and random sampling. Then, we performed SLEPA on 14-layer graphene-WS2 heterostructures. Moreover, we extracted three features which could suppress phonon transmission across the full range of frequency and angle of incidence. Finally, we constructed an empirical model which could predict thermal conductivity of graphene-WS2 heterostructure with 70% accuracy.
|
