“…Set-to-sequence ML models also play an important role in data-intensive 3D point cloud processing (Qi, Su, Mo, & Guibas, 2017) and meta-learning (Huang, Wang, Singh, Yih, & He, 2018b). Set-input and set-ordering problems themselves are prominent in a wide array of applications ranging from power grid optimization (Cui, Yu, Li, Zeng, & Gu, 2019), where solving them led to power usage savings of up to 30%, through anomaly detection (Jung, Berges, Garrett, & Poczos, 2015) to measurements of contaminated galaxy clusters (Ntampaka, Trac, Sutherland, Fromenteau, Poczos, & Schneider, 2016).…”