“…In many real-world applications, obtaining labeled samples is very costly, while a large number of unlabeled samples are readily available. To exploit unlabeled samples and improve the accuracies of learners, AL and SSL have been extensively investigated for many real-world problems in machine learning, such as text classification (Tong and Chang, 2001;Hoi et al, 2006;Burkhardt et al, 2018), information extraction (Thompson et al, 1999;Wu and Pottenger, 2005), image classification and retrieval (Hoi and Lyu, 2005;Li et al, 2013;Pedronette et al, 2019), and cancer diagnosis (Nguyen et al, 2020;Menon et al, 2020).…”