| Abstract Scope |
Artificial intelligence (AI) techniques were explored as tools for imbalance learning, inverse design and transfer learning for high entropy alloys (HEAs). In the first case study, the minority phase data of HEAs were augmented using generative models, such as variational autoencoder (VAE), generative adversarial network (GAN), and denoising diffusion probabilistic model (DDPM), to address the data imbalance issue, and their performance was compared. In the second case study, the DDPM was assessed as a potential tool for the inverse design of HEAs. A DDPM-assisted generative inverse design framework was proposed, and its efficient compositional optimization was demonstrated. In the last case study, a transfer learning technique was leveraged to address the data scarcity of additively manufactured (AMed) HEAs. A basis model for the hardness prediction was pre-trained using a comprehensive dataset of as-cast HEAs, and fine-tuned to predict the hardness of AMed HEAs. |