Targeted Background Removal Creates Interpretable Feature Visualizations
This work addresses interpretability challenges for researchers and practitioners using black-box models, though it appears incremental as it builds on existing feature visualization methods with a specific training modification.
The paper tackles the problem of improving interpretability in feature visualizations for black-box machine learning models by using background removal techniques during training, resulting in visualizations that show significant improvement with easily recognizable features compared to baseline models.
Feature visualization is used to visualize learned features for black box machine learning models. Our approach explores an altered training process to improve interpretability of the visualizations. We argue that by using background removal techniques as a form of robust training, a network is forced to learn more human recognizable features, namely, by focusing on the main object of interest without any distractions from the background. Four different training methods were used to verify this hypothesis. The first used unmodified pictures. The second used a black background. The third utilized Gaussian noise as the background. The fourth approach employed a mix of background removed images and unmodified images. The feature visualization results show that the background removed images reveal a significant improvement over the baseline model. These new results displayed easily recognizable features from their respective classes, unlike the model trained on unmodified data.