Graph-based Heuristic Search for Module Selection Procedure in Neural Module Network
This work addresses a training bottleneck in NMNs for visual question answering, offering a more efficient alternative to existing methods, though it appears incremental in scope.
The paper tackles the non-differentiable module selection problem in Neural Module Networks (NMNs) for visual question answering by proposing a Graph-based Heuristic Search algorithm, which trains NMNs without ground-truth programs and achieves superior efficiency over reinforcement learning methods on FigureQA and CLEVR datasets.
Neural Module Network (NMN) is a machine learning model for solving the visual question answering tasks. NMN uses programs to encode modules' structures, and its modularized architecture enables it to solve logical problems more reasonably. However, because of the non-differentiable procedure of module selection, NMN is hard to be trained end-to-end. To overcome this problem, existing work either included ground-truth program into training data or applied reinforcement learning to explore the program. However, both of these methods still have weaknesses. In consideration of this, we proposed a new learning framework for NMN. Graph-based Heuristic Search is the algorithm we proposed to discover the optimal program through a heuristic search on the data structure named Program Graph. Our experiments on FigureQA and CLEVR dataset show that our methods can realize the training of NMN without ground-truth programs and achieve superior efficiency over existing reinforcement learning methods in program exploration.