Latent Space Explanation by Intervention
This work addresses interpretability in deep learning, particularly for identifying and mitigating bias in datasets, though it appears incremental as it builds on existing intervention and visualization techniques.
The study tackled the problem of interpreting deep neural networks by revealing hidden concepts that drive predictions, using an intervention mechanism with discrete variational autoencoders to shift predicted classes and visualize changes, demonstrating effectiveness on the CelebA dataset to show and alter data bias.
The success of deep neural nets heavily relies on their ability to encode complex relations between their input and their output. While this property serves to fit the training data well, it also obscures the mechanism that drives prediction. This study aims to reveal hidden concepts by employing an intervention mechanism that shifts the predicted class based on discrete variational autoencoders. An explanatory model then visualizes the encoded information from any hidden layer and its corresponding intervened representation. By the assessment of differences between the original representation and the intervened representation, one can determine the concepts that can alter the class, hence providing interpretability. We demonstrate the effectiveness of our approach on CelebA, where we show various visualizations for bias in the data and suggest different interventions to reveal and change bias.