“…Thus using the attributes of the data is critical for the transfer learning (Kulkarni, Sharma, Zepeda and Chevallier, 2014;Peng, Tian, Xiang, Wang, Pontil and Huang, 2017;Suzuki, Sato, Oyama and Kurihara, 2014a,b). In this paper, we study the problem of effective use of both input data and attribute for the domain-transfer learning problem, and propose a novel method of attribute embedding based on the popular convolutional neural network (CNN) (Fujino, Hatanaka, Mori and Matsumoto, 2018;Geng, Liang, Li, Wang, Liang, Xu and Wang, 2016;Geng, Zhang, Li, Gu, Liang, Liang, Wang, Wu, Patil and Wang, 2017;Jing, Zhao, Li and Xu, 2017;Puri, Tewari, Katyal and Garg, 2018;Roa-Barco, Serradilla-Casado, Velasco-Vzquez, Lpez-Zorrilla, Gra?a, Chyzhyk and Price, 2018;Shen, Zhou, Yang, Yang and Tian, 2015;Shen, Zhou, Yang, Yu, Dong, Yang, Zang and Tian, 2017;Todoroki, Han, Iwamoto, Lin, Hu and Chen, 2018;Waijanya and Promrit, 2018;Zhang, Liang, Li, Fang, Wang, Geng and Wang, 2017a) to solve this problem. Further, we develop a novel model using the attribute embedding as the input for the learning of the target domain classification model.…”