This research examines how model explanations change under distribution shifts in tabular data.
problem Detecting distribution shifts in tabular data affecting model performance and explanations.
method Investigates the relationship between model performance and explanation characteristics under distribution shifts.
result Explanation shifts are a better indicator for detecting predictive performance changes than traditional distribution shift techniques.
CTE improves explanation estimation with less data and faster computation.
problem Inefficient and inaccurate explanation estimation in machine learning models.
method Distribution compression through kernel thinning to reduce sample size.
result CTE significantly improves accuracy and stability of explanation estimation.
LLMs' explanations are often insufficient and vary with input distribution.
problem Evaluating the sufficiency of LLM explanations without predefined biases.
method Generalizing sufficiency to arbitrary explanations, using LLM's input beliefs, and introducing SCSuff metric.
result Explanation sufficiency can vary with input distribution and is weakly correlated with model size, accuracy, or output entropy.
Extends local attributions to Bayesian Neural Networks for improved explanations.
problem Lack of explanations for Bayesian Neural Networks' predictions.
method Extend local attributions to a probabilistic explanation distribution of BNNs.
result Enriches standard explanations with uncertainty information and visualizes explanation stability.
New approach models how explanations shift with distribution changes.
problem Model performance drops with changing input data distributions.
method Models explanation shifts and compares them to state-of-the-art techniques.
result Modeling explanation shifts better detects out-of-distribution behavior.
DistShap parallelizes GNN explanation for large graphs.
problem Computational expense in attributing GNN predictions to specific edges or features.
method Distributed Shapley values across multiple GPUs for scalable GNN explanations.
result DistShap outperforms existing methods and scales to models with millions of features.
This paper introduces DCE for better counterfactual explanations using optimal transport.
problem Lack of nuanced distributional characteristics in existing counterfactual explanations.
method Formulates a chance-constrained optimization problem using optimal transport to derive counterfactual distributions.
result DCE provides deeper insights into decision-making models by aligning counterfactual distributions with factual ones.
ManifoldShap improves model explanations by restricting evaluations to the data manifold.
problem Inaccurate and misleading model explanations due to reliance on out-of-distribution data.
method Restricts model evaluations to the data manifold to avoid off-manifold perturbations.
result ManifoldShap provides more accurate and intuitive explanations than existing methods.
PACE explains ViTs by modeling patch-level concept distributions, surpassing existing methods.
problem Lack of trustworthy post-hoc explanations for Vision Transformers (ViTs)
method Variational Bayesian explanation framework (PACE)
result PACE surpasses state-of-the-art methods in meeting desiderata for ViT explanations.
GRANITE unifies feature-based explanation methods to reduce disagreement.
problem Disagreement among feature-based explanation methods.
method GRANITE partitions feature space into regions minimizing interaction and distribution influences.
result Unified and consistent feature explanations.
GLIME improves LIME's stability and local fidelity.
problem LIME's instability and low local fidelity.
method Introducing GLIME, an enhanced framework that derives an equivalent formulation of LIME with faster convergence and improved stability.
result GLIME generates explanations with higher local fidelity and is independent of reference choice.
New framework quantifies uncertainties in neural network explanations.
problem Lack of methods to quantify uncertainties in neural network explanations.
method Converts any explanation method into a Bayesian neural network method, modeling uncertainties.
result Allows quantification of explanation uncertainties and appropriate confidence levels.
We consider objective evaluation measures of saliency explanations for complex black-box machine learning models. We propose simple robust variants of two notions that have been considered in recent literature: (in)fidelity, and sensitivity. We analyze optimal explanations with respect to both these measures, and while…
RAW-Explainer generates interpretable subgraph explanations for link predictions in knowledge graphs.
problem Interpreting GNN predictions for link prediction in heterogeneous settings is challenging.
method RAW-Explainer uses random walk objective and neural network to generate connected, concise subgraph explanations.
result RAW-Explainer strikes a balance between explanation quality and computational efficiency.
The study evaluates how well local explanations align with model predictions.
problem Capturing the faithfulness of local explanations to model predictions.
method Introducing consistency and sufficiency as properties, and developing quantitative measures and estimators.
result Quantitative measures of consistency and sufficiency depend on test-time data distribution.
This paper analyzes errors in Shapley value-based model explanations.
problem Biased or unreliable explanations from existing SVA methods.
method Error theoretical analysis framework decomposing errors into observation and structural biases.
result Trade-off between observation and structural biases in SVA explanations.
Symmetric observations don't necessarily imply symmetric causal explanations.
problem Inferring causal models from observed correlations is challenging and computationally intensive.
method An explicit example using a tripartite probability distribution over binary events.
result Symmetries in observations cannot be used to reduce the hypothesis space of causal models.
We introduce a method, KL-LIME, for explaining predictions of Bayesian predictive models by projecting the information in the predictive distribution locally to a simpler, interpretable explanation model. The proposed approach combines the recent Local Interpretable Model-agnostic Explanations (LIME) method with ideas …
EBLIME enhances model explanations using Bayesian ridge regression.
problem Improving model explanations for black-box machine learning models.
method EBLIME uses Bayesian ridge regression to explain feature importance.
result EBLIME provides more intuitive and accurate feature importance rankings.
Neural networks memorize exceptions, leading to poor generalization.
problem Memorization of exceptions hinders neural network generalization.
method Formalized memorization-generalization interplay, proposed MAT to shift logits.
result MAT improves generalization by learning robust patterns invariant across distributions.
New metric assesses reliability of AI explanations.
problem Unreliable AI explanations under realistic conditions.
method Explanation Reliability Index (ERI) metrics quantifying stability under four axioms.
result Widespread reliability failures in popular explanation methods.
BIF assesses feature importance using Dirichlet distribution and Bayesian inference.
problem Quantitative feature importance assessment in statistical models.
method Utilizes Dirichlet distribution for probabilistic feature importance assessment via approximate Bayesian inference.
result Learned importance provides relative significance and confidence quantification of features.
Develops causal explanations for black-box classifiers.
problem Creating understandable explanations for complex models.
method Generative model with information-theoretic measures of causal influence.
result Learned latent factors influence classifier outputs.
MCCE generates realistic counterfactual explanations for tabular data.
problem Creating valid and actionable counterfactual explanations for complex tabular data.
method MCCE models the joint distribution of features and decision using an autoregressive generative model with decision trees. It samples counterfactuals and removes invalid ones.
result MCCE outperforms state-of-the-art methods on various performance metrics and is faster.
As machine learning black boxes are increasingly being deployed in domains such as healthcare and criminal justice, there is growing emphasis on building tools and techniques for explaining these black boxes in an interpretable manner. Such explanations are being leveraged by domain experts to diagnose systematic error…
Unified framework for feature-based explanations using ANOVA and game theory.
problem Differences between feature-based explanations methods limit their applicability.
method Introduces a unified framework combining fANOVA and cooperative game theory.
result Uncovered similarities and differences between various explanation techniques.
Language helps RL agents learn complex relational and causal structures.
problem Learning relational and causal structure in complex environments.
method Training RL agents to predict language descriptions and explanations.
result Language aids agents in learning challenging relational and causal tasks.
Proposes simplified SHAP for faster black-box model explanations.
problem Computational expense of SHAP for models with many features.
method Ensemble of random SHAPs with feature selection and point generation.
result Efficiency and properties demonstrated through numerical experiments.
We extend and improve the work of Model Agnostic Anchors for explanations on image classification through the use of generative adversarial networks (GANs). Using GANs, we generate samples from a more realistic perturbation distribution, by optimizing under a lower dimensional latent space. This increases the trust in …
Identification of disease subtypes and corresponding biomarkers can substantially improve clinical diagnosis and treatment selection. Discovering these subtypes in noisy, high dimensional biomedical data is often impossible for humans and challenging for machines. We introduce a new approach to facilitate the discovery…
ALMANACS benchmarks explainability methods on simulatability.
problem Evaluating the effectiveness of explainability methods for language models.
method ALMANACS is a simulatability benchmark that evaluates explainability methods on twelve safety-relevant topics.
result No explainability method outperforms the explanation-free control across all topics.
Model explanations based on pure observational data cannot compute the effects of features reliably, due to their inability to estimate how each factor alteration could affect the rest. We argue that explanations should be based on the causal model of the data and the derived intervened causal models, that represent th…
An imprecise SHAP method explains class probabilities with limited data.
problem Explaining class probabilities with limited training data.
method New approach for computing feature marginal contributions and general approach to interval-valued Shapley values.
result The imprecise SHAP method improves explanation of class probabilities.
New method uses SHapley Additive Explanations to identify anomaly detectors with complementary behaviors.
problem Challenges in unsupervised anomaly detection due to diverse data distributions and lack of labels.
method Characterize anomaly detectors using SHapley Additive Explanations to measure feature importance and similarity.
result Detectors with similar explanations produce correlated anomaly scores, while those with divergent explanations are complementary.
Paper introduces a method to explain concept drift using counterfactual explanations.
problem Understanding the features where concept drift occurs for better model adjustment.
method Formal definition and algorithm based on counterfactual explanations.
result Demonstrates usefulness of the method in various examples.
Proposes counterfactual explanations for deep two-sample tests on high-dimensional data.
problem Limited interpretability of deep two-sample tests on high-dimensional data.
method Combines diffusion autoencoder and pretrained deep two-sample test model to generate counterfactuals.
result Counterfactual transformations increase p-values, indicating closer distribution similarity.
Proposes a method for clearer counterfactual explanations of deep networks.
problem Unclear explanations from current counterfactual methods.
method Gradual construction of explanations through masking and composition steps.
result Produces human-friendly, interpretable explanations with fewer modifications.
Generative models explain machine learning predictions with counterfactual instances.
problem Generating human-interpretable insights into machine learning model predictions.
method Sparse counterfactual explanations using conditional generative models.
result Single forward pass generates batches of counterfactual instances.
Defines globalness measure for explainers using optimal transport.
problem Challenges in evaluating and comparing explainability methods.
method Axiomatic definition and proof of Wasserstein Globalness measure.
result Wasserstein Globalness measure facilitates meaningful comparison and selection of explainers.
Method generates counterfactual explanations for graph classifiers.
problem Generating high-quality explanations for graph predictions.
method Permutation equivariant graph variational autoencoder to traverse latent space.
result Empirically validated model is high-performing and robust.
In simulations of some economic gas-like models, the asymptotic regime shows an exponential wealth distribution, independently of the initial wealth distribution given to the system. The appearance of this statistical equilibrium for this type of gas-like models is explained in a rigorous analytical way.
OptiLIME improves LIME explanations by balancing stability and adherence.
problem LIME's instability and lack of reliability in explanations.
method OptiLIME uses a deterministic sampling approach and feature selection to maximize stability while retaining predefined adherence.
result OptiLIME provides more reliable and interpretable explanations.
Game-theoretic formulations of feature importance have become popular as a way to "explain" machine learning models. These methods define a cooperative game between the features of a model and distribute influence among these input elements using some form of the game's unique Shapley values. Justification for these me…
We study the problem of explaining a rich class of behavioral properties of deep neural networks. Distinctively, our influence-directed explanations approach this problem by peering inside the network to identify neurons with high influence on a quantity and distribution of interest, using an axiomatically-justified in…
Many methods to explain black-box models, whether local or global, are additive. In this paper, we study global additive explanations for non-additive models, focusing on four explanation methods: partial dependence, Shapley explanations adapted to a global setting, distilled additive explanations, and gradient-based e…
The paper covers the new model of wage distribution in typical group of people. The model provides the opportunity to reparameterize applicable income distribution model: Pareto, logarithmically normal, logarithmically logistic, Dagum etc. The model ensures the graduation of Gini index values by polynomial degree of wa…
New methods identify concepts in trained embeddings reliably without human labels.
problem Identifying interpretable concepts in trained embedding spaces without human labels.
method Explicitly connecting concept discovery to PCA and ICA, proposing novel approaches for dependent concepts.
result Proven methods outperform competitors on a variety of experiments, achieving up to 29% better alignment with ground truth.
AXE evaluates explanations to avoid misleading Rashomon set model selection.
problem Evaluating explanations for Rashomon set models to avoid false selection.
method Proposed AXE method to evaluate explanation quality.
result AXE detects adversarial fairwashing with 100% success rate.