Visual Explanations from Hadamard Product in Multimodal Deep Networks

TL;DR

This paper demonstrates that the Hadamard product in multimodal deep networks acts as an implicit attention mechanism for both visual and textual inputs, and introduces a gradient-based visualization method to analyze this effect.

Contribution

It extends previous work by showing Hadamard product's attentional role for both modalities and proposes a new visualization technique for interpretation.

Findings

Hadamard product implicitly performs attention for visual inputs.

The method visualizes attention for both visual and textual inputs.

Comparison with learned attention weights validates the approach.

Abstract

The visual explanation of learned representation of models helps to understand the fundamentals of learning. The attentional models of previous works used to visualize the attended regions over an image or text using their learned weights to confirm their intended mechanism. Kim et al. (2016) show that the Hadamard product in multimodal deep networks, which is well-known for the joint function of visual question answering tasks, implicitly performs an attentional mechanism for visual inputs. In this work, we extend their work to show that the Hadamard product in multimodal deep networks performs not only for visual inputs but also for textual inputs simultaneously using the proposed gradient-based visualization technique. The attentional effect of Hadamard product is visualized for both visual and textual inputs by analyzing the two inputs and an output of the Hadamard product with the proposed method and compared with learned attentional weights of a visual question answering model.