When Explanations Lie: Why Modified BP Attribution Fails

by   Leon Sixt, et al.
Freie Universität Berlin

Modified backpropagation methods are a popular group of attribution methods. We analyse the most prominent methods: Deep Taylor Decomposition, Layer-wise Relevance Propagation, Excitation BP, PatternAttribution, Deconv, and Guided BP. We found empirically that the explanations of the mentioned modified BP methods are independent of the parameters of later layers and show that the z^+ rule used by multiple methods converges to a rank-1 matrix. This can explain well why the actual network's decision is ignored. We also develop a new metric cosine similarity convergence (CSC) to directly quantify the convergence of the modified BP methods to a rank-1 matrix. Our conclusion is that many modified BP methods do not explain the predictions of deep neural networks faithfully.


page 4

page 12


An induction proof of the backpropagation algorithm in matrix notation

Backpropagation (BP) is a core component of the contemporary deep learni...

Electricity consumption forecasting method based on MPSO-BP neural network model

This paper deals with the problem of the electricity consumption forecas...

Exact Reconstruction Conditions for Regularized Modified Basis Pursuit

In this correspondence, we obtain exact recovery conditions for regulari...

Understanding Individual Decisions of CNNs via Contrastive Backpropagation

A number of backpropagation-based approaches such as DeConvNets, vanilla...

Improving Massive MIMO Belief Propagation Detector with Deep Neural Network

In this paper, deep neural network (DNN) is utilized to improve the beli...

Exploiting the Relationship Between Kendall's Rank Correlation and Cosine Similarity for Attribution Protection

Model attributions are important in deep neural networks as they aid pra...

Deep Layer-wise Networks Have Closed-Form Weights

There is currently a debate within the neuroscience community over the l...