How can we fool LIME and SHAP? Adversarial Attacks on Post hoc Explanation Methods.
Dylan SlackSophie HilgardEmily JiaSameer SinghHimabindu LakkarajuPublished in: CoRR (2019)
Keyphrases
- empirical studies
- post hoc
- significant improvement
- qualitative and quantitative
- learning algorithm
- computational cost
- neural network
- decision trees
- database
- evolutionary algorithm
- feature space
- statistical methods
- computationally expensive
- machine learning methods
- benchmark datasets
- preprocessing
- case study
- feature selection
- computer vision
- data sets
- real time