1. R. Geirhos, P. Rubisch, C. Michaelis, M. Bethge, F. Wichmann, W. Brendel, ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness, arXiv:abs/1811.12231(2019).
2. Dual learning of the generator and recognizer for chinese characters;Zhu,2017
3. Radical composition network for chinese character generation;Xue,2021
4. Attention is all you need;Vaswani,2017
5. Multi-agent dual learning;Wang,2019