Explaining Classifications For Individual Instances

By Marko Robnik-Sikonja, Igor Kononenko. 2008

293 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Safety, Robustness & Interpretability

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

In Computer Vision and Pattern Recognition

In Safety, Robustness & Interpretability

Citations in the corpus

Explanation in Human-AI Systems: A Literature Meta-Review, Synopsis of Key Ideas and Publications, and Bibliography for Explainable AI (2019, neutral): “Možina, Demšar, Bratko, and Zupan, 2005; Robnik-Šikonja and Kononenko, 2008; Robnik-Šikonja, Likas, Constantinopoulos, Kononenko, and Štrumbelj, 2011).”
Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI) (2018, methodology): “[93]–[95]).”
Explanations of model predictions with live and breakDown packages (2018, background): “It is worth noting that similar decomposition of predictions and measures of contribution for classifiers have been studied in Robnik-Šikonja and Kononenko (2008). The greedy search can start from null set of indexes (then in each step a single feature is being relaxed) or it can start from full set”
Explanation Methods in Deep Learning: Users, Values, Concerns and Challenges (2018, background): “Attribution, a term introduced by [2], also referred to as relevance [5, 7, 60, 47], contribution [51], class saliency [52] or influence [27, 1, 28], aims to reveal components of high importance in the input to the DNN and their effect as the input is propagated through the network.”
Explanation in Artificial Intelligence: Insights from the Social Sciences (2017, background): “looked at selecting which features in the model were important for a decision, mostly built on local explanations [157, 6, 156] or on information gain [90, 89].”
Enhancing Transparency and Control When Drawing Data-Driven Inferences About Individuals (2016, background): “phenomenon that we want to provide transparency into.(22) Explaining instance-level classifications by assessing the (causal) influence of data inputs is”
How to Explain Individual Classification Decisions (2009, methodology): “To save the time of combinatorial training Robnik-ˇSikonja and Kononenko (2008) propose to use neutral values which have to be estimated by a known prior distribution of all possible parameter values.”
Visualizing Deep Neural Network Decisions: Prediction Difference Analysis (2017, methodology): “The work presented in this paper is based on an instance-speciﬁc method by Robnik-Šikonja & Kononenko (2008), the prediction difference analysis , which is reviewed in the next section.”
Detecting concept drift in data streams using model explanation (2018, methodology): “E P T Several related explanation methods that study the inﬂuence of each individual attribute on its own with limitations in considering their potentially intertwined relations exist in the related work (Robnik-ˇSikonja & Kononenko, 2008; Lemaire et al., 2008; Lughofer et al., 2016a).”
Regional Multi-Scale Approach for Visually Pleasing Explanations of Deep Neural Networks (2018, methodology): “Robnik-Sikonja and Kononenko proposed the prediction differencemethod [16], and several studies have extended and improved this method [15], [17].”
Explainable artificial intelligence (XAI) in deep learning-based medical image analysis (2021, methodology): “(2017) adapted prediction difference analysis (Robnik-Šikonja and Kononenko (2008)) for generating saliency maps. If each pixel in an image is considered a feature, prediction difference analysis assigns a relevance value to each pixel, by measuring how the prediction changes if the pixel is conside”
Explaining Data-Driven Document Classifications (2014, methodology): “Specifically, a procedure for producing explanations of document classifications may provide no explanation at all.”
Explaining prediction models and individual predictions with feature contributions (2014, background): “Previous general approaches [19,24] tackled the problem in the following way:”
A New Method to Visualize Deep Neural Networks (2016, methodology): “We want to build upon the method of Robnik-ˇSikonja and Kononenko (2008) to develop a tool for analyzing how DC-NNs classify images.”
Explaining by Removing: A Unified Framework for Model Explanation (2020, methodology): “, 2017), which is based on a precursor method to IME (Robnik-Šikonja and Kononenko, 2008; Štrumbelj et al., 2009), removes individual features (or groups of features) and analyzes the difference in the model’s prediction.”
Efficiently Explaining Decisions of Probabilistic RBF Classification Networks (2011, methodology): “We first present the simpler of the two [6] (called EXPLAIN in the reminder of the text), which computes the influence of the feature value by changing this value and observing its impact on the model’s output.”
Uncertainty Estimation in Bayesian Neural Networks And Links to Interpretability (2018, methodology): “[69] is multivariate analysis, rather than univariate analysis in [53].”
Justification Narratives for Individual Classifications (2014, background): “…the model in a human-consumable form (Szafron et al., 2003) or as the problem of projecting the processes and parameters of non-linear models to make them more meaningful to the expert (Robnik-Sikonja and Kononenko, 2008; Baehrens et al., 2010), while justification as a goal has not been mentioned.”