Explanations based on the Missing: Towards Contrastive Explanations with Pertinent Negatives

Authors: Amit Dhurandhar, Pin-Yu Chen, Ronny Luss, Chun-Chen Tu, Paishun Ting, Karthikeyan Shanmugam, Payel Das

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this paper we propose a novel method that provides contrastive explanations justifying the classification of an input by a black box classifier such as a deep neural network. ... We validate our approach on three real datasets obtained from diverse domains; namely, a handwritten digits dataset MNIST, a large procurement fraud dataset and a brain activity strength dataset. In all three cases, we witness the power of our approach in generating precise explanations that are also easy for human experts to understand and evaluate.
Researcher Affiliation Collaboration Amit Dhurandhar IBM Research Yorktown Heights, NY 10598 adhuran@us.ibm.com; Pin-Yu Chen IBM Research Yorktown Heights, NY 10598 pin-yu.chen@ibm.com; Ronny Luss IBM Research Yorktown Heights, NY 10598 rluss@us.ibm.com; Chun-Chen Tu University of Michigan Ann Arbor, MI 48109 timtu@umich.edu; Paishun Ting University of Michigan Ann Arbor, MI 48109 paishun@umich.edu; Karthikeyan Shanmugam IBM Research Yorktown Heights, NY 10598 karthikeyan.shanmugam2@ibm.com; Payel Das IBM Research Yorktown Heights, NY 10598 daspa@us.ibm.com
Pseudocode Yes Algorithm 1 Contrastive Explanations Method (CEM)
Open Source Code Yes Code at https://github.com/IBM/Contrastive-Explanation-Method
Open Datasets Yes a handwritten digits dataset MNIST [40]... a brain functional MRI (f MRI) imaging dataset from the publicly accessible Autism Brain Imaging Data Exchange (ABIDE) I database [11]
Dataset Splits Yes The handwritten digits are classified using a feed-forward convolutional neural network (CNN) trained on 60,000 training images from the MNIST benchmark dataset. ... The 10-fold cross validation accuracy of the network was high (91.6%). ... The leave-one-out cross validation testing accuracy is around 61.17%.
Hardware Specification No The paper does not provide specific hardware details such as GPU models, CPU types, or memory amounts used for running experiments.
Software Dependencies No The paper mentions 'Tensor Flow' but does not specify a version number. It also refers to 'LRP experiments used the toolbox from [21]' and 'LIME code was adapted from https://github.com/marcotcr/lime' without providing specific software versions for these tools.
Experiment Setup Yes The parameter κ 0 is a confidence parameter that controls the separation between [Pred(x0 + δ)]t0 and maxi =t0[Pred(x0 + δ)]i. The parameters c, β, γ, 0 are the associated regularization coefficients. We apply a projected fast iterative shrinkage-thresholding algorithm (FISTA) [2] to solve problems (1) and (3). The CNN has two sets of convolution-convolution-pooling layers, followed by three fully-connected layers. We trained a three-layer neural network with fully connected layers, 512 rectified linear units and a three-way softmax function. The parameters of the model were regularized by an elastic-net regularizer.