Beyond saliency: Understanding convolutional neural networks from saliency prediction on layer-wise relevance propagation

Abstract

Despite the tremendous achievements of deep convolutional neural networks (CNNs) in many computer vision tasks, understanding how they actually work remains a significant challenge. In this paper, we propose a novel two-step understanding method, namely Salient Relevance (SR) map, which aims to shed light on how deep CNNs recognize images and learn features from areas, referred to as attention areas, therein. Our proposed method starts out with a layer-wise relevance propagation (LRP) step which estimates a pixelwise relevance map over the input image. Following, we construct a contextaware saliency map, SR map, from the LRP-generated map which predicts areas close to the foci of attention instead of isolated pixels that LRP reveals. In human visual system, information of regions is more important than of pixels in recognition. Consequently, our proposed approach closely simulates human recognition. Experimental results using the ILSVRC2012 validation dataset in conjunction with two well-established deep CNN models, AlexNet and VGG-16, clearly demonstrate that our proposed approach concisely identifies not only key pixels but also attention areas that contribute to the underlying neural network's comprehension of the given images. As such, our proposed SR map constitutes a convenient visual interface which unveils the visual attention of the network and reveals which type of objects the model has learned to recognize after training. The source code is available at https://github.com/Hey1Li/Salient-Relevance-Propagation.

Secure and Robust Machine Learning for Healthcare: A Survey

“…In [203], the use of reinforcement learning (RL) is proposed to build interpretable decision support systems for heart patients and it learns what is interpretable to each user by their interactions. One yet common method for interpreting/explaining deep models, in particular, CNN is the use of saliency maps [204], [205]. These methods are particularly focused on general applications, however, more research that is specifically focused on the interpretation of ML/DL systems used in healthcare applications is required.…”

Visual Analytics in Deep Learning: An Interrogative Survey for the Next Frontiers

“…One research area generates translucent heatmaps that overlay images to highlight important regions that contribute towards classification and their sensitivity [4], [69], [92], [93], [94]. One technique called visual backpropagation attempts to visualize which parts of an image have contributed to the classification, and can do so in real-time in a model debugging tool for self-driving vehicles [30].…”

Benchmarking and Survey of Explanation Methods for Black Box Models

“…A variant of -lrp is spray [80] which builds a specrtal clustering on top of the local instance-based -lrp explanations. Similar work is done in [82]: it starts with the -lrp of the input instance and finds the LRP attribution relevance for a single input of interest x.…”