Bias Also Matters: Bias Attribution for Deep Neural Network Explanation

Authors: Shengjie Wang, Tianyi Zhou, Jeff Bilmes

ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In experiments, we show that BBp can generate complementary and highly interpretable explanations.In Table 1, we compare results on the CIFAR-10, CIFAR-100 (Krizhevsky & Hinton, 2009) and Fashion MNIST (Xiao et al., 2017) datasets.
Researcher Affiliation Academia 1Paul G. Allen School of Computer Science & Engineering 2Department of Electrical & Computer Engineering, University of Washington, Seattle, USA.
Pseudocode Yes Algorithm 1 Bias Backpropagation (BBp)
Open Source Code No The paper does not provide any explicit statement about open-source code availability or a link to a code repository for the described methodology.
Open Datasets Yes We compare results on the CIFAR-10, CIFAR-100 (Krizhevsky & Hinton, 2009) and Fashion MNIST (Xiao et al., 2017) datasets.We test BBp on STL-10 (Coates et al., 2011) and Image Net(ILSVRC2012) (Russakovsky et al., 2015).
Dataset Splits No In Table 1, we compare results on the CIFAR-10, CIFAR-100 (Krizhevsky & Hinton, 2009) and Fashion MNIST (Xiao et al., 2017) datasets.
Hardware Specification No The paper does not provide specific details on the hardware used for running experiments, such as exact GPU or CPU models.
Software Dependencies No The paper does not specify particular software dependencies with version numbers, such as specific deep learning frameworks or libraries.
Experiment Setup Yes For STL-10, we use a 10-layer convolutional network ((32,3,1), maxpool, (64,3,1), maxpool, (64,3,1), maxpool, (128,3,1), maxpool, (128,3,1), and dense10, where (i,j,k) corresponds to a convolutional layer with i channels, kernel size j and padding k), and for Image Net we use the VGG-11 network of (Simonyan & Zisserman, 2014).For all options of calculating the bias attribution score, the temperature parameter T is set to 1.