The Bayesian Case Model: A Generative Approach for Case-Based Reasoning and Prototype Classification

By Been Kim, Cynthia Rudin, Julie Arnold Shah. 2014

341 citations recorded by the source corpus.

We present the Bayesian Case Model (BCM), a general framework for Bayesian case-based reasoning (CBR) and prototype classification and clustering. BCM brings the intuitive power of CBR to a Bayesian generative framework. The BCM learns prototypes, the "quintessential" observations that best represent clusters in a dataset, by performing joint inference on cluster labels, prototypes and important features. Simultaneously, BCM pursues sparsity by learning subspaces, the sets of features that play important roles in the characterization of the prototypes. The prototype and subspace representation provides quantitative benefits in interpretability while preserving classification accuracy. Human subject experiments verify statistically significant improvements to participants' understanding when using explanations produced by BCM, compared to those given by prior art.

Subfields: Artificial Intelligence, Probabilistic & Bayesian Methods

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 Probabilistic & Bayesian Methods

Citations in the corpus

SELFEXPLAIN: A Self-Explaining Architecture for Neural Text Classifiers (2021, methodology): “…in the training data for already-trained models, inherently interpretable models (Melis and Jaakkola, 2018; Arik and Pﬁs-ter, 2020) may provide greater transparency since explanation capability is embedded directly within the model (Kim et al., 2014; Doshi-Velez and Kim, 2017; Rudin, 2019).”
Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges (2021, background): “[146] formulated a prototype-parts learning problem for structured (tabular) data using a Bayesian generative framework.”
Deep k-Nearest Neighbors: Towards Confident, Interpretable and Robust Deep Learning (2018, methodology): “follow-ups [58], [59], [60], this technique was also applied to visualize relationships learned between words by the word2vec language model [61].”
Interpreting Deep Classifier by Visual Distillation of Dark Knowledge (2018, background): “…to interpret (Craven & Shavlik, 1996), retrieving typical inputs so that one can interpret by examples via hidden activations (Caruana et al., 1999) or inﬂuence functions (Koh & Liang, 2017) and generating artifacts or prototypes to explain model behaviour (Ribeiro et al., 2016; Kim et al., 2015).”
Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV) (2017, methodology): “Examples are shown to be effective means of interfacing with ML models for both non-expert and expert users (Koh & Liang, 2017; Kim et al., 2014; 2015; Klein, 1989).”
Nothing Else Matters: Model-Agnostic Explanations By Identifying Prediction Invariance (2016, background): “Experiments usually involve showing humans these visualizations, and measuring human precision when predicting the model’s behavior on random instances, and the time (effort) required to make those predictions [7, 8, 9].”
The Dangers of Post-hoc Interpretability: Unjustified Counterfactual Explanations (2019, methodology): “Other methods rely on using speciﬁc instances to build explanations using comparison to relevant neighbors, such as prototype-based approaches [Kim et al. , 2014] and counterfactual explanation approaches.”
Deep Weighted Averaging Classifiers (2018, background): “The explanation is thus given in terms of exemplars or prototypes , which have been shown to be an effective approach to interpretability [18].”
This looks like that: deep learning for interpretable image recognition (2018, methodology): “Our work relates closely to other prototype classification techniques in machine learning [32, 2, 45], and very closely, to the Bayesian Case Model [18].”
Interpretable Decision Sets: A Joint Framework for Description and Prediction (2016, background): “For example, prototype vector machines [4], regression models [36, 35], generalized additive models [16, 40, 26, 27], Bayesian case models [19] mind-the-gap models [20], and scoring systems [37] have all been proposed as interpretable models.”
Machine Unlearning (2019, background): “exist is that when each constituent model is trained on very limited data which is also not prototypical [30]—especially when the number of samples per class is low; if the model has high-capacity (as is the case with DNNs), the model might overfit to the small training dataset.”
ConvNets and ImageNet Beyond Accuracy: Explanations, Bias Detection, Adversarial Examples and Model Criticism (2017, methodology): “They have extensively been used in case-based reasoning (CBR) to improve the interpretability of the models [1, 4, 25].”
Explaining Explanations in AI (2018, methodology): “Case-based explanation methods (Caruana et al., 1999; Kim et al., 2014) for non-cased based machine learning involve using the trained model as a distance metric to determine which cases in the training data set are most similar to the case or decision to be explained.”