Inducing Sparse Coding and And-Or Grammar from Generator Network
This work addresses the need for interpretable generative models in computer vision, though it appears incremental as it builds on existing generator network frameworks.
The paper tackles the problem of learning explainable hierarchical representations in generative models by applying sparse operations on generator network feature maps, resulting in a method that induces both sparse coding and AND-OR grammar for images.
We introduce an explainable generative model by applying sparse operation on the feature maps of the generator network. Meaningful hierarchical representations are obtained using the proposed generative model with sparse activations. The convolutional kernels from the bottom layer to the top layer of the generator network can learn primitives such as edges and colors, object parts, and whole objects layer by layer. From the perspective of the generator network, we propose a method for inducing both sparse coding and the AND-OR grammar for images. Experiments show that our method is capable of learning meaningful and explainable hierarchical representations.