Neural Module Networks
This addresses the challenge of combining deep networks with compositional reasoning for visual question answering, representing a novel method rather than an incremental improvement.
The paper tackled the problem of visual question answering by exploiting the compositional nature of questions, using neural module networks that dynamically instantiate modular components based on linguistic substructures, achieving state-of-the-art results on the VQA natural image dataset and a new abstract shapes dataset.
Visual question answering is fundamentally compositional in nature---a question like "where is the dog?" shares substructure with questions like "what color is the dog?" and "where is the cat?" This paper seeks to simultaneously exploit the representational capacity of deep networks and the compositional linguistic structure of questions. We describe a procedure for constructing and learning *neural module networks*, which compose collections of jointly-trained neural "modules" into deep networks for question answering. Our approach decomposes questions into their linguistic substructures, and uses these structures to dynamically instantiate modular networks (with reusable components for recognizing dogs, classifying colors, etc.). The resulting compound networks are jointly trained. We evaluate our approach on two challenging datasets for visual question answering, achieving state-of-the-art results on both the VQA natural image dataset and a new dataset of complex questions about abstract shapes.