based method ◦ エントロピー(不確かさ)最大のデータを選ぶ • Least Confident ◦ ラベルの確率の最大値が最小になるデータを選択 1. Uncertainty Sampling 12 y_i: 各クラスのラベル A B C D score x1 0.10 0.10 0.10 0.70 0.70 x2 0.25 0.25 0.25 0.25 0.25 A B C D score x1 0.10 0.10 0.10 0.70 0.94 x2 0.25 0.25 0.25 0.25 1.39
2009, Burr Settles, Computer Sciences Technical Report 1648 • Active Learning 入門, https://www.slideshare.net/shuyo/introduction-to-active-learning-25787487 • Overview of Active Learning for Deep Learning, https://jacobgil.github.io/deeplearning/activelearning#active-learning-for-convolution al-neural-networks--a-core-set-approach • Active Learning for Convolutional Neural Networks: A Core-Set Approach, 2018, Ozan Sener, ICLR • Variational Adversarial Active Learning, 2019, Samarth Sinha, ICCV • BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning, 2019, Andreas Kirsch, NeurIPS • 能動学習:問題設定と最近の話題, 2021, 日野英逸, 日本統計学会誌 参考