Hierarchical Dynamic Masks for Visual Explanation of Neural Networks
This work addresses the need for more detailed and robust explanations of neural network classifications, particularly in domains like medical imaging, though it is incremental as it builds on existing saliency methods.
The authors tackled the problem of generating visual explanatory maps for neural network decisions by proposing Hierarchical Dynamic Masks (HDM), which enhanced granularity and comprehensiveness, resulting in significant outperformance in recognition and localization capabilities on natural and medical datasets.
Saliency methods generating visual explanatory maps representing the importance of image pixels for model classification is a popular technique for explaining neural network decisions. Hierarchical dynamic masks (HDM), a novel explanatory maps generation method, is proposed in this paper to enhance the granularity and comprehensiveness of saliency maps. First, we suggest the dynamic masks (DM), which enables multiple small-sized benchmark mask vectors to roughly learn the critical information in the image through an optimization method. Then the benchmark mask vectors guide the learning of large-sized auxiliary mask vectors so that their superimposed mask can accurately learn fine-grained pixel importance information and reduce the sensitivity to adversarial perturbations. In addition, we construct the HDM by concatenating DM modules. These DM modules are used to find and fuse the regions of interest in the remaining neural network classification decisions in the mask image in a learning-based way. Since HDM forces DM to perform importance analysis in different areas, it makes the fused saliency map more comprehensive. The proposed method outperformed previous approaches significantly in terms of recognition and localization capabilities when tested on natural and medical datasets.