Understanding classifier errors by examining influential neighbors

By Mayank Kabra, Alice Robie, Kristin Branson. 2015

29 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Auditing black-box models for indirect influence (2016, neutral): “visualizing the behavior of neurons and inputs have also been studied [28], [14], [16].”
FairTest: Discovering Unwarranted Associations in Data-Driven Applications (2015, methodology): “A number of methods exist in ML to understand data features and modeling errors [33]–[36].”
Auditing Black-box Models by Obscuring Features (2016, background): “at understanding how these make decisions on image inputs have included visualizing the derived representation by reversing a neural net to output an image when given a label [17], probing nodes to determine what stimulus image produces the highest activation [12], and examining activation patterns ”
Inverse Classification for Comparison-based Interpretability in Machine Learning (2017, background): “For instance, Kabra et al. (2015) try to identify which train observations have the highest direct inﬂuence over a single prediction.”
Discovering Unwarranted Associations in Data-Driven Applications with the FairTest Testing Toolkit (2015, methodology): “Fourth, debugging the ML component in an application is often conducted in an ad hoc fashion, with the ML practitioner manually employing exploratory data analysis to understand data features and modeling errors [24, 29, 40, 41].”
Comparison-Based Inverse Classification for Interpretability in Machine Learning (2018, methodology): “Instance-based Approaches Instance-based approaches constitute a family of post-hoc methods that bring interpretability by comparing an observation to relevant neighbors [22, 13, 1].”
TIDE: A General Toolbox for Identifying Object Detection Errors (2020, background): “Many works exist to analyze the errors in object detection and instance segmentation [15,24,7,17,22], but only a few provide a useful summary of all the errors in a model [14,1,4], and none have all the desirable properties listed above.”
A Survey on the Explainability of Supervised Machine Learning (2020, methodology): “Here, we also demand explanations for why the model did not perform well on these few feature combinations (Kabra et al., 2015).”