Gradient-based methods improve understanding of deep learning survival models.
problem Limited interpretability of deep learning survival models hinders their adoption.
method Gradient-based explanation methods tailored to survival neural networks.
result Gradient-based methods capture feature effects and temporal dynamics.
Gradient-based explanations correlate with Android malware classifier robustness.
problem Evasion attacks on Android malware classifiers using sparse perturbations.
method Investigated gradient-based attribution methods for explaining classifier decisions and their evenness, proposing metrics to assess adversarial robustness.
result Gradient-based explanations, especially Integrated Gradients, correlate with adversarial robustness of malware classifiers.
WassersteinGrad improves weather forecasting explanations by addressing geometric misalignment issues.
problem Improving explainability of autoregressive neural predictions on dynamic physical fields.
method WassersteinGrad, a geometric consensus method for averaged perturbed attribution maps.
result WassersteinGrad provides more accurate explanations for weather forecasting models.
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…
A technique scales symbolic methods with gradients for neural model explanation.
problem Limited scalability of symbolic methods for large neural networks.
method Combines gradient-based methods with symbolic techniques using Integrated Gradients to focus on a subset of neurons.
result Produces sparser and higher saliency regions compared to gradient-based methods alone.
Method generates visual explanations for similarity models without classification.
problem Lack of visual explanations for similarity models trained without classification loss.
method Gradient-based visual attention using learned feature embeddings.
result Attention maps improve model performance and can be used as constraints.
Survey of counterfactual explanations for time series classification.
problem Generating plausible and actionable counterfactuals for time series data.
method Review of various counterfactual generation methods for time series classification.
result Highlight unique challenges and strengths of existing methods.
EXAGREE selects a stakeholder-aligned model to reduce conflicting explanations in machine learning.
problem Conflicting explanations from different attribution methods limit the adoption of machine learning models in safety-critical domains.
method EXAGREE is a two-stage framework that selects a Stakeholder-Aligned Explanation Model (SAEM) from a set of similar-performing models, maximizing Stakeholder-Machine Agreement (SMA).
result EXAGREE achieves simultaneous gains in faithfulness, plausibility, and fairness over baselines while preserving task accuracy.
ID-ExpO fine-tunes neural networks for more faithful explanations.
problem Improving the faithfulness of explanations for complex machine learning models.
method Differentiable insertion/deletion metric-aware regularizers for optimization.
result Fine-tuned predictors produce more faithful explanations.
We propose a novel perspective to understand deep neural networks in an interpretable disentanglement form. For each semantic class, we extract a class-specific functional subnetwork from the original full model, with compressed structure while maintaining comparable prediction performance. The structure representation…
RELAX provides first attribution-based explanations for representations.
problem Lack of methods to explain what influences learned representations.
method RELAX, a first approach for attribution-based explanations of representations, measuring similarities in representation space.
result Significantly outperforms gradient-based baseline and models uncertainty in explanations.
We show through theory and experiment that gradient-based explanations of a model quickly reveal the model itself. Our results speak to a tension between the desire to keep a proprietary model secret and the ability to offer model explanations. On the theoretical side, we give an algorithm that provably learns a two-la…
Mode connectivity is a surprising phenomenon in the loss landscape of deep nets. Optima -- at least those discovered by gradient-based optimization -- turn out to be connected by simple paths on which the loss function is almost constant. Often, these paths can be chosen to be piece-wise linear, with as few as two segm…
PatternLocal improves XAI for non-linear models by suppressing suppressor variables.
problem Suppressor variables cause false-positive feature attributions in non-linear models.
method PatternLocal uses locally linear surrogate models and transforms weights into a generative representation.
result PatternLocal reduces false-positive attributions and provides more reliable explanations.
New method makes neural network explanations more robust to attacks.
problem Vulnerability of feature attributions to adversarial attacks.
method Promotes Lipschitz continuity and smoothness in DNNs to improve robustness.
result Regularization and stochastic smoothing methods enhance attribution robustness.
MoreauGrad interprets neural nets robustly and sparsely.
problem Lack of robustness and sparsity in gradient-based interpretation methods.
method Moreau envelope for smooth and robust interpretation, combined with L1 regularization for sparsity.
result MoreauGrad provides a smooth, robust, and sparse explanation of neural nets.
Our work investigates disagreement in neural network feature attribution methods.
problem Disagreement among feature attribution methods for neural networks.
method Investigates the fundamental and distributional behavior of feature attribution methods.
result Illustrates the impact of scaling and encoding techniques on explanation quality.
DeepUnHide uses deep learning to reveal hidden demographic features in recommender systems.
problem Extracting hidden demographic features from recommender systems factors.
method Gradient-based localization in deep learning for feature extraction.
result DeepUnHide outperforms state-of-the-art feature selection methods.
MACQ method explains deep learning models by analyzing feature contributions across prediction levels.
problem Explaining deep learning model predictions.
method Global gradient-based, model-agnostic approach focusing on marginal attribution.
result MACQ separates feature contributions from interaction effects and visualizes 3-way relationships.
New method generates plausible counterfactuals for time series classification.
problem Generating realistic counterfactuals for time series data.
method Gradient-based optimization with soft-DTW alignment and multi-faceted loss function.
result Our method outperforms existing approaches in temporal realism and distributional alignment.
Recently, a technique called Layer-wise Relevance Propagation (LRP) was shown to deliver insightful explanations in the form of input space relevances for understanding feed-forward neural network classification decisions. In the present work, we extend the usage of LRP to recurrent neural networks. We propose a specif…
Explaining the output of a deep network remains a challenge. In the case of an image classifier, one type of explanation is to identify pixels that strongly influence the final decision. A starting point for this strategy is the gradient of the class score function with respect to the input image. This gradient can be …
The paper explains how language models acquire complex skills through scaling laws and statistical analysis.
problem Understanding how language models acquire new skills as their size and training data increase.
method Statistical framework and mathematical analysis of scaling laws.
result Language models can learn complex skills efficiently due to a strong inductive bias.
New probabilistic approaches offer recourse recommendations even when causal models are imperfect.
problem Limited causal knowledge makes guaranteeing algorithmic recourse impossible.
method Two probabilistic approaches: Bayesian model averaging and average effect computation.
result Probabilistic approaches lead to more reliable recourse recommendations.
Recently, a spate of papers have provided positive theoretical results for training over-parameterized neural networks (where the network size is larger than what is needed to achieve low error). The key insight is that with sufficient over-parameterization, gradient-based methods will implicitly leave some components …
As Convolutional Neural Networks embed themselves into our everyday lives, the need for them to be interpretable increases. However, there is often a trade-off between methods that are efficient to compute but produce an explanation that is difficult to interpret, and those that are slow to compute but provide a more i…
Method identifies root causes of anomalies in causal processes.
problem Identifying root causes of anomalies in causal processes.
method Noisy functional causal model, Bayesian learning, gradient-based attribution.
result Proposes efficient method to compute anomaly attribution scores.
This paper reviews zeroth-order optimization in signal processing and machine learning.
problem Optimization problems without gradient information.
method Iterative steps: gradient estimation, descent direction computation, solution update.
result Demonstrates applications in robustness evaluation and black-box model explanations.
Rank concepts help explain deep learning's effectiveness.
problem Limited formal understanding of deep learning.
method Introducing rank concepts to explain generalization and expressiveness.
result Gradient-based training induces low rank in neural networks.
New methods improve deep learning for solving linear PDEs.
problem Efficiently solving high-dimensional linear PDEs using deep learning.
method Rigorous investigation of gradient estimators for SDE-based variational formulations.
result Novel methods provide substantial performance improvements.
CDAM improves attention maps for ViTs, making them more class-sensitive.
problem Existing attention maps in ViTs lack class sensitivity.
method Class-discriminative attention maps (CDAM) that scale attention scores by class relevance.
result CDAM provides more class-sensitive explanations than existing methods.
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.
The paper proposes criteria and methods for evaluating and aggregating feature-based model explanations.
problem Lack of quantitative evaluation criteria for feature-based model explanations.
method Developed quantitative evaluation criteria (low sensitivity, high faithfulness, low complexity), devised a framework for aggregation, and derived a new aggregate Shapley value explanation function.
result A new aggregate Shapley value explanation function that minimizes sensitivity.
New definition reveals encoding explanations that retain predictive power.
problem Challenges in evaluating and identifying encoding explanations.
method Developed a definition of encoding based on conditional dependence.
result Existing evaluation scores do not rank non-encoding explanations correctly, but STRIPE-X does.
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.
Differentially private algorithms protect model explanations from leaking training data.
problem Model explanations can leak training data, compromising privacy.
method Adaptive differentially private gradient descent algorithm to produce accurate, private explanations.
result Privacy amplification and reduction of overall privacy loss on explanation data.
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…
The selection of initial parameter values for gradient-based optimization of deep neural networks is one of the most impactful hyperparameter choices in deep learning systems, affecting both convergence times and model performance. Yet despite significant empirical and theoretical analysis, relatively little has been p…
Introduces new gradient-based methods for machine learning problems.
problem New challenges in machine learning due to decision-making and multi-agent problems.
method Gradient-based optimization and variational inequalities.
result Shifts focus from pattern recognition to decision-making and multi-agent problems.
Improves global counterfactual explanations for model recourse.
problem Inability to provide explanations beyond local instances.
method Investigates and improves Actionable Recourse Summaries (AReS) for global counterfactual explanations.
result Develops more efficient and interactive explainability tools.
New framework evaluates model explanations based on decision task improvement.
problem Evaluation of model explanations often misses practical value.
method Decision-theoretic framework quantifying three key values.
result Provides benchmarks and interprets human-AI decision support.
We define and compute plausible counterfactual explanations using density constraints.
problem Efficiently compute plausible counterfactual explanations for machine learning models.
method Propose and study a formal definition of plausible counterfactual explanations, use density estimators, and introduce convex density constraints.
result Convex density constraints ensure plausible and feasible counterfactual explanations.
G-SHAP generates multiple types of explanations for machine learning models.
problem Understanding model predictions and their differences across groups.
method Generalization of SHAP method to produce additional types of explanations.
result G-SHAP produces explanations for classification, intergroup differences, and model failure.
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.
Study finds visual explanations do not significantly improve human accuracy or trust in model predictions.
problem Measuring the impact of visual explanations on human accuracy and trust in model predictions.
method Randomized controlled trial with image-based age prediction task, varying levels of explanation quality.
result Visual explanations do not significantly alter human accuracy or trust in the model.
Despite a growing literature on explaining neural networks, no consensus has been reached on how to explain a neural network decision or how to evaluate an explanation. Our contributions in this paper are twofold. First, we investigate schemes to combine explanation methods and reduce model uncertainty to obtain a sing…
Bayesian Neural Networks are robust to gradient-based attacks in the large-data limit.
problem Vulnerability of deep learning models to adversarial attacks.
method Analysis of adversarial attacks in the large-data, overparametrized limit for Bayesian Neural Networks.
result BNN posteriors are robust to gradient-based adversarial attacks in the limit.
Many proposed methods for explaining machine learning predictions are in fact challenging to understand for nontechnical consumers. This paper builds upon an alternative consumer-driven approach called TED that asks for explanations to be provided in training data, along with target labels. Using semi-synthetic data fr…