Theano implementation of the CSDNN model proposed in Cost-Aware Pre-Training for Multiclass Cost-Sensitive Deep Learning, in Proceedings of the 25th International Joint Conference on Artificial Intelligence (IJCAI), 2016.
If you use the code, please cite the paper as:
@inproceedings{chung2016csdnn,
title = {Cost-aware pre-training for multiclass cost-sensitive deep learning},
author = {Chung, Yu-An and Lin, Hsuan-Tien and Yang, Shao-Wen},
booktitle = {IJCAI},
year = {2016}
}
There's also a follow-up work here: Cost-Sensitive Deep Learning with Layer-Wise Cost Estimation
@article{chung2016auxcst,
title = {Cost-sensitive deep learning with layer-wise cost estimation},
author = {Chung, Yu-An and Lin, Hsuan-Tien},
journal = {arXiv preprint arXiv:1611.05134},
year = {2016}
}