Spatial Dependency Networks: Neural Layers for Improved Generative Image Modeling
This work addresses the challenge of better exploiting spatial regularities in images for generative modeling, particularly in VAEs, offering a domain-specific improvement for image generation tasks.
The authors tackled the problem of improving generative image modeling by introducing spatial dependency networks (SDNs), a novel neural network layer that enhances spatial coherence in image generators. They applied SDNs to variational autoencoders (VAEs) and achieved improved density estimation over baseline convolutional architectures and state-of-the-art models in the same class, with demonstrations of high-quality sample synthesis on large images.
How to improve generative modeling by better exploiting spatial regularities and coherence in images? We introduce a novel neural network for building image generators (decoders) and apply it to variational autoencoders (VAEs). In our spatial dependency networks (SDNs), feature maps at each level of a deep neural net are computed in a spatially coherent way, using a sequential gating-based mechanism that distributes contextual information across 2-D space. We show that augmenting the decoder of a hierarchical VAE by spatial dependency layers considerably improves density estimation over baseline convolutional architectures and the state-of-the-art among the models within the same class. Furthermore, we demonstrate that SDN can be applied to large images by synthesizing samples of high quality and coherence. In a vanilla VAE setting, we find that a powerful SDN decoder also improves learning disentangled representations, indicating that neural architectures play an important role in this task. Our results suggest favoring spatial dependency over convolutional layers in various VAE settings. The accompanying source code is given at https://github.com/djordjemila/sdn.