Domain Separation Networks.

Friday. October 16, 2020 - 1 min

Domain Separation Networks.

[NIPS2016] Bousmalis, K., Trigeorgis, G., Silberman, N., Krishnan, D., & Erhan, D. (Google Brain, Imperial College London, Google Research). Domain separation networks. In Advances in neural information processing systems (pp. 343-351).

简介

这篇文章介绍的域独立网络，通过学习域不变表达，来解决域适应问题。

相关工作

寻找源域到目标域的映射（26）
Sun, B., Feng, J., & Saenko, K. (2015). Return of frustratingly easy domain adaptation. arXiv preprint arXiv:1511.05547.
寻找域不变的表达（8,28,17）
Ganin, Y., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., … & Lempitsky, V. (2016). Domain-adversarial training of neural networks. The Journal of Machine Learning Research, 17(1), 2096-2030.
Tzeng, E., Hoffman, J., Darrell, T., & Saenko, K. (2015). Simultaneous deep transfer across domains and tasks. In Proceedings of the IEEE International Conference on Computer Vision (pp. 4068-4076).
Long, M., Cao, Y., Wang, J., & Jordan, M. (2015, June). Learning transferable features with deep adaptation networks. In International conference on machine learning (pp. 97-105). PMLR.
尺度不变平均平方损失（6）
Eigen, D., Puhrsch, C., & Fergus, R. (2014). Depth map prediction from a single image using a multi-scale deep network. In Advances in neural information processing systems (pp. 2366-2374).

方法

本文介绍的网络结构示意图如下：

图1 网络结构示意图
该网络包括5个部分：首先是独立地对源域和目标域的图像进行编码，得到特征$h_p^s$和$h_p^t$，然后是共享的编码器去学习域不变特征$h_c^s$和$h_c^t$，并利用这些特征进行源域和目标域图像的复原，得到$\hat{x}^s$和$\hat{x}^s$，当然也少不了分类网络对源域图像进行分类。总的Loss是： $\mathscr{L} = \mathscr{L}_{task} + \alpha\mathscr{L}_{recon} + \beta\mathscr{L}_{difference} + \gamma\mathscr{L}_{similarity}$ 其中$\mathscr{L}_{task}$为任务的loss，对于分类任务，则自然为交叉熵loss： $\mathscr{L}_{task} = - \sum_{i=0}^{N_s}y_i^slog\hat{y}_i^s$ 重建的loss采用尺度不变平均平方损失，为： $\mathscr{L}_{recon} = \sum_{i=1}^{N_s}\mathscr{L}_{si\_mse}(x_i^s,\hat{x}_i^s)+\sum_{i=1}^{N_t}\mathscr{L}_{si\_mse}(x_i^t,\hat{x}_i^t)$ 其中： $\mathscr{L}_{si\_mse}(x,\hat{x}) = \frac{1}{k}||x-\hat{x}||_2^2-\frac{1}{k^2}((x-\hat{x})\cdot 1_k)^2$ 差异的Loss则是特征级别的，鼓励独立的特征和共享的特征尽可能正交，即： $\mathscr{L}_{difference} = ||(H_c^s)^TH_p^s||_F^2 + ||(H_c^t)^TH_p^t||_F^2$ 相似性loss包括两部分，其中一部分类似于对抗网络，有： $\mathscr{L}_{similarity}^{DANN} = \sum_{i=0}^{N_s+N_t}{d_ilog\hat{d}_i+(1-d_i)log(1-\hat{d}_i)}$ 另一部分Loss为特征级别的基于核方法的最大差异Loss，为 $\mathscr{L}_{similarity}^{MMD} = \frac{1}{(N^s)^2}\sum^{N^s}_{i,j=0}\kappa(h^s_{ci},h^s_{cj})-\frac{2}{N^sN^t}\sum^{N^s,N^t}_{i,j=0}\kappa(h^s_{ci},h^t_{cj})+\frac{1}{(N^t)^2}\sum^{N^t}_{i,j=0}\kappa(h^t_{ci},h^t_{cj})$