Equivariant Transformer Networks
Equivariant Transformer Networks
复制标题
DOI：
--
复制
发表时间：
2019-01
期刊：
影响因子：
--
通讯作者：
Kai Sheng Tai;Peter D. Bailis;G. Valiant
中科院分区：
文献类型：
--
作者：
Kai Sheng Tai;Peter D. Bailis;G. Valiant
如何将有关域转换不变的先验知识纳入神经网络的架构中？我们提出了epoiriant变形金刚（ETS），这是一个由图像到图像映射的家族，可以改善模型对预定的连续变换组的鲁棒性。通过使用特殊的规范坐标系，ETS结合了相对于这些转换而构造的函数。我们从经验上表明，可以灵活地组成ET，以提高对更复杂的转换组的模型鲁棒性。在现实世界图像分类任务上，ETS提高了重新网络分类器的样本效率，在有限的数据制度中，错误率相对提高了高达15％的错误率，同时将模型参数计数提高了不到1％。
How can prior knowledge on the transformation invariances of a domain be incorporated into the architecture of a neural network? We propose Equivariant Transformers (ETs), a family of differentiable image-to-image mappings that improve the robustness of models towards pre-defined continuous transformation groups. Through the use of specially-derived canonical coordinate systems, ETs incorporate functions that are equivariant by construction with respect to these transformations. We show empirically that ETs can be flexibly composed to improve model robustness towards more complicated transformation groups in several parameters. On a real-world image classification task, ETs improve the sample efficiency of ResNet classifiers, achieving relative improvements in error rate of up to 15% in the limited data regime while increasing model parameter count by less than 1%.