3:30 PM - 3:45 PM
▲ [11p-N107-8] Search for suitable hyperparameters in Bayesian optimization for material synthesis: three-dimensional case
Keywords:Machine learning, Bayesian Optimization
Conventional materials exploration requires trial-and-error in multi-dimensional searching space. However, in general, it is difficult for humans to optimize multi-dimensional synthesis conditions. Bayesian optimization has aroused the attention of material researchers as an efficient method to explore novel functional materials. In order to find the optimal material synthesis conditions with the minimum number of experiments, we have already investigated the optimum hyperparameters of the kernel function and the acquisition function in Bayesian optimization using 1D model functions. However, the optimum hyperparameters in the 3D case are still not investigated. In this study, we determined the appropriate hyperparameters that cost the least number of experiments for optimization in 3D synthesis conditions through simulation, and compared the effect of the process window on searching speed.