This paper proposes the decision tree latent controller generative adversarial network (DTLC-GAN), an extension of a GAN that can learn hierarchically interpretable representations without relying on detailed supervision. To impose a hierarchical inclusion structure on latent variables, we incorporate a new architecture called the DTLC into the generator input. The DTLC has a multiple-layer tree structure in which the ON or OFF of the child node codes is controlled by the parent node codes. By using this architecture hierarchically, we can obtain the latent space in which the lower layer codes are selectively used depending on the higher layer ones. To make the latent codes capture salient semantic features of images in a hierarchically disentangled manner in the DTLC, we also propose a hierarchical conditional mutual information regularization and optimize it with a newly defined curriculum learning method that we propose as well. This makes it possible to discover hierarchically interpretable representations in a layer-by-layer manner on the basis of information gain by only using a single DTLC-GAN model. We evaluated the DTLC-GAN on various datasets, i.e., MNIST, CIFAR-10, Tiny ImageNet, 3D Faces, and CelebA, and confirmed that the DTLC-GAN can learn hierarchically interpretable representations with either unsupervised or weakly supervised settings. Furthermore, we applied the DTLC-GAN to image-retrieval tasks and showed its effectiveness in representation learning.
翻译:本文建议了决定树潜控制器变基因对抗网络(DTLC-GAN),这是GAN的延伸,可以在不依赖详细监督的情况下学习分等级解释的表达方式。为了在潜在变量上强行设置一个等级包容结构,我们将一个称为DTLC的新结构纳入生成器输入中。DTLC有一个多层树结构,在这种结构中,儿童节点的 On 或 FFF 由父节点代码控制。通过使用这一结构,我们可以获得一个潜在的空间,根据较高层次来选择使用低层代码。为了使潜在代码在DTLCLC中以等级分解的方式捕捉到图像的突出的语义性特征,为了在DTLCLC-G中以等级分解,我们还提出一个等级分级的相互信息规范,并用我们提议的新定义的课程学习方法加以优化。这样就可以在信息收益的基础上,通过使用单一的DTLC-G-G-GAN模型,我们评估了DLC-G-LC-G-S-DLC-S-ILC-S-ILC-LC-ILC-S-ILV 和S-S-ILV-S-S-ILV-ILV-S-C-C-C-C-C-C-C-C-ILVD-ILV 和S-S-ID-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-ILVILVDLVD-C-ILVD-C-C-C-C-C-C-C-C-C-C-I-C-C-C-C-D-ILV-I-I-I-I-I-I-I-I-I-C-I-I-