A post-hoc framework improves model performance by calibrating different feature spaces.
problem Improving AUC performance on binary classification tasks for overconfident models.
method Identifies heterogeneous partitions of the feature space and applies post-hoc calibration techniques to each partition.
result Theoretical optimality of the framework for any model, demonstrated on deep neural networks.
Post-hoc model-agnostic interpretation methods such as partial dependence plots can be employed to interpret complex machine learning models. While these interpretation methods can be applied regardless of model complexity, they can produce misleading and verbose results if the model is too complex, especially w.r.t. f…
EAGLE improves reproducibility and stability of model explanations.
problem Creating reliable explanations for opaque machine learning models.
method Formulates perturbation selection as an information-theoretic active learning problem.
result EAGLE learns a linear surrogate model with feature importance scores and uncertainty estimates.
Study compares various calibration methods for binary classification tasks.
problem Improving probabilistic predictions in binary classification models.
method Benchmarked 21 classifiers using 5 calibration methods on real data.
result Venn-Abers predictors and Beta calibration show the largest log-loss reductions.
Most of the work on interpretable machine learning has focused on designing either inherently interpretable models, which typically trade-off accuracy for interpretability, or post-hoc explanation systems, whose explanation quality can be unpredictable. Our method, ExpO, is a hybridization of these approaches that regu…
TaylorPODA uses Taylor expansions to improve feature attributions for opaque models.
problem Lack of systematic framework for quantifying feature contributions in opaque models.
method Taylor expansion framework with postulates (precision, federation, zero-discrepancy, adaptation).
result TaylorPODA achieves competitive results and provides principled explanations.
XCM improves MTS classification with explainable deep learning.
problem Lack of explainable deep learning models for MTS classification.
method XCM is a compact CNN that extracts variable and timestamp information directly from input data.
result XCM outperforms state-of-the-art MTS classifiers on large and small datasets.
Surrogate explainers of black-box machine learning predictions are of paramount importance in the field of eXplainable Artificial Intelligence since they can be applied to any type of data (images, text and tabular), are model-agnostic and are post-hoc (i.e., can be retrofitted). The Local Interpretable Model-agnostic …
Enhances GNN robustness during inference using Conditional Random Fields.
problem Vulnerability of GNNs to adversarial attacks.
method Post-hoc approach using Conditional Random Fields (CRF).
result Improves robustness of GNNs across various models.
Proposes MOC method for better counterfactual explanations in ML models.
problem Difficulties in balancing multiple objectives for counterfactual explanations.
method Translates counterfactual search into a multi-objective optimization problem.
result Returns diverse counterfactuals with different trade-offs and maintains feature diversity.
A new OOD detection method OTOD uses optimal transport theory to improve model performance.
problem Detecting unknown samples in real-world machine learning models.
method OTOD uses optimal transport theory to calculate an OOD score combining features, logits, and softmax probability space.
result OTOD outperforms state-of-the-art methods by significant margins on benchmarks.
DECAT framework evaluates multimodal models for shared biology, detecting confounders and false positives.
problem Determining if multimodal models learn shared biology or just confounders.
method DECAT framework classifies multimodal representations into four diagnostic scenarios using null-referenced metrics.
result DECAT detects confounders and false positives in multimodal models, improving with larger cohorts and stronger representations.
Proposes a method to balance fairness and utility in ranking models.
problem Systematic disparity across protected groups in ranking models.
method Model-agnostic post-processing framework using dynamic programming.
result Achieves a balance between fairness and utility across various metrics and datasets.
CDLEEDS detects local changes in evolving data streams for accurate feature attributions.
problem Local feature attributions become obsolete in evolving data streams.
method CDLEEDS, a flexible framework for detecting local change and concept drift.
result CDLEEDS reliably detects both local and global concept drift.
This paper enhances credit risk management using explainable AI techniques.
problem Lack of transparency and explainability in AI models for credit risk management.
method Implement LIME and SHAP for explaining ML-based credit scoring models.
result Demonstrates practical challenges and solutions for XAI methods in finance.
Post-hoc transforms can reverse model performance trends, especially in noisy settings.
problem Post-hoc transforms can reverse model performance trends, especially in noisy settings.
method Empirical study and analysis of post-hoc transforms like temperature scaling, ensembling, and SWA.
result Post-hoc reversal can prevent double descent and mitigate mismatches between test loss and test error.
Post-hoc calibration improves uncertainty under domain shift.
problem Improving uncertainty calibration under domain shift.
method Apply perturbations to validation set before post-hoc calibration.
result Perturbation step results in better calibration under domain shift.
Researchers show how to manipulate Partial Dependence plots to deceive explanations of predictive models.
problem The robustness and trustworthiness of Partial Dependence (PD) explanations are compromised.
method Data poisoning using genetic and gradient algorithms to manipulate PD plots.
result PD explanations can be fooled and manipulated to mislead understanding of predictive models.
FFCP improves FCP's speed without sacrificing accuracy.
problem Inefficient feature transformation in FCP.
method Introduces FFCP using Taylor expansion for faster computation.
result FFCP achieves a 50x speedup with comparable accuracy.
Risk Advisor predicts and mitigates ML deployment failures.
problem Predicting and mitigating test-time failure risks of ML systems.
method Post-hoc meta-learner for estimating failure risks and uncertainties.
result Reliably predicts deployment-time failure risks across various ML models.
Proposes DFDG for robust domain generalization without source domain labels.
problem Robustness of deep learning models in real-world applications where train and test distributions differ.
method Model-agnostic, class-aware alignment of class relationships through saliency maps.
result Competitive performance on time series sensor and image classification datasets.
Framework evaluates post-hoc interpretability methods in time-series classification.
problem Lack of suitable post-hoc interpretability methods for time-series classification.
method Proposes a framework with quantitative metrics to assess interpretability methods.
result Addresses several drawbacks of existing methods, including dependence on human judgement and data distribution shift.
Mathematical study shows post-hoc explanations are better than attention weights alone.
problem Understanding the internal behavior of attention-based models.
method Mathematical analysis of a simple attention-based architecture.
result Post-hoc explanations provide more useful insights than attention weights alone.
CalArena benchmarks post-hoc calibration methods across various tasks.
problem Inconsistent evaluations of post-hoc calibration methods.
method Large-scale benchmark with 2000 experiments, covering diverse models and settings.
result Smooth calibration functions outperform binning-based approaches.
Post hoc test for Sharpe ratio improves pairwise comparisons.
problem Improving pairwise comparisons of Sharpe ratios.
method Analogous to Tukey's test, applied after rejecting equal Signal-Noise ratios.
result Maintains nominal type I rate and is moderately powerful.
Meta-Cal improves post-hoc calibration of neural networks.
problem Improving the accuracy of uncalibrated neural network predictions.
method Meta-Cal uses a base calibrator and a ranking model with constraints to provide high-probability bounds.
result Meta-Cal significantly outperforms existing methods in post-hoc multi-class classification calibration.
New approach interprets machine learning models through feature space transformations.
problem Interpreting models with strongly dependent features in high-dimensional spaces.
method Feature space transformations, including PCA and partial orthogonalization.
result Enhances model interpretation tools for domain experts.
Developed an explainable DRL model for financial portfolio management.
problem Inability of DRL agents to provide interpretable financial investment policies.
method Integrating PPO with feature importance techniques (SHAP, LIME) to enhance transparency.
result Ability to interpret DRL agent actions in prediction time.
DBPA assesses LLM perturbations using frequentist hypothesis testing.
problem Quantifying input perturbation impacts on LLM outputs.
method DBPA reformulates perturbation analysis as frequentist hypothesis testing, using Monte Carlo sampling for empirical null and alternative distributions.
result DBPA provides interpretable p-values and scalar effect sizes for LLM perturbations.
Study post-hoc Learning to Defer using density-ratio losses.
problem Optimizing decision-making between models and experts.
method Density-ratio losses for post-hoc L2D scorers, derived from class-probability estimation.
result The approach recovers known results and introduces new connections to expert comparison and anomaly detection.
Post-hoc calibration of neural networks using g-Layers proves theoretical justification.
problem Ensuring the confidence of neural network decisions in real-world applications.
method Proves theoretical justification for post-hoc calibration methods by adding g-Layers and minimizing NLL.
result Proves that adding g-Layers and minimizing NLL can lead to a calibrated network.
End-to-end method improves neural network calibration during training.
problem Improving neural network calibration for regression problems.
method Quantile Recalibration Training integrates post-hoc calibration into model training.
result Improved predictive accuracy and calibration in a large-scale experiment.
Counterfactual post-hoc interpretability approaches have been proven to be useful tools to generate explanations for the predictions of a trained blackbox classifier. However, the assumptions they make about the data and the classifier make them unreliable in many contexts. In this paper, we discuss three desirable pro…
A new method calibrates value predictions in offline RL to improve reliability.
problem Difficulty in long-horizon value prediction in offline reinforcement learning.
method Bellman calibration, a weak reliability criterion, and Iterated Bellman Calibration.
result Finite-sample guarantees show that Bellman calibration error is controlled at nonparametric rates.
Paper improves deep learning for instance-level classification from label proportions.
problem Dealing with noisy pseudo-labeling and high-entropy class distributions in LLP.
method Introducing a two-stage training approach with constrained optimization and mixup strategy.
result Significant performance improvement in instance-level classification.
Anchors explains text classifiers by highlighting key words.
problem Interpreting machine learning models, especially for text classifiers.
method Formalizes Anchors' algorithm and analyzes its behavior on linear text classifiers.
result Anchors produces meaningful results on linear text classifiers.
The paper compares ML models for credit scoring and investment decisions using explainable AI.
problem The opacity of machine learning models in financial services.
method Comparison of various machine learning models (single classifiers, ensembles, neural networks) and explainability techniques (LIME, SHAP).
result Ensemble classifiers and neural networks outperform in credit scoring models.
Building an open-domain conversational agent is a challenging problem. Current evaluation methods, mostly post-hoc judgments of static conversation, do not capture conversation quality in a realistic interactive context. In this paper, we investigate interactive human evaluation and provide evidence for its necessity; …
LIMEtree offers faithful explanations for multiple classes in predictive models.
problem Generating explanations for several classes can be difficult due to conflicting evidence.
method LIMEtree uses multi-output regression trees for consistent and faithful explanations of multiple classes.
result LIMEtree provides diverse explanation types and outperforms LIME in evaluations.
New method calibrates deep networks by preserving top-k predictions.
problem Calibrated confidence scores for multi-class deep networks to avoid rare mistakes.
method Intra order-preserving functions combined with neural network architecture.
result Outperforms state-of-the-art methods in evaluation metrics.
While statistics and machine learning offers numerous methods for ensuring generalization, these methods often fail in the presence of adaptivity---the common practice in which the choice of analysis depends on previous interactions with the same dataset. A recent line of work has introduced powerful, general purpose a…
Model agnostic feature importance for DNNs in NLP.
problem Characterizing DNNs as black boxes and explaining their decision-making process.
method Phrase-wise feature importance calculation for model agnostic DNNs.
result Robust and generalizable approach to feature importance calculation.
Post-hoc explanations improve CNNs by replacing final linear layer with k-means classifier.
problem CNNs lack accurate data representation in their built-in prototypes.
method Introduces k-means-based post-hoc explanations for CNNs, leveraging spatial consistency of convolutional receptive fields.
result Using shallower, less compressed feature activations improves semantic fidelity at the cost of slight predictive performance.
A novel post-hoc calibration method reduces neural network calibration errors.
problem Neural networks produce poorly calibrated probabilities, leading to underconfidence and overconfidence.
method Probability bounding (PB) via box-constrained softmax (BCSoftmax) function.
result Consistently reduces calibration errors on four real-world datasets.
AI methods broaden signal discovery in scientific data.
problem Limited coverage of possible signals in model-dependent searches.
method Model-agnostic AI strategies for broad exploration.
result Enhanced discovery potential in experimental science.
Model-agnostic interpretation techniques allow us to explain the behavior of any predictive model. Due to different notations and terminology, it is difficult to see how they are related. A unified view on these methods has been missing. We present the generalized SIPA (sampling, intervention, prediction, aggregation) …
A new method for measuring conditional feature importance using generative models.
problem Challenges in evaluating feature importance given other feature values.
method Adversarial Random Forest (ARF) for generating on-manifold data points.
result cARFi method yields robust importance scores adaptable for various feature importance notions.
COCKATIEL explains neural net models on NLP tasks by identifying meaningful concepts.
problem Transformer models are complex and hard to interpret.
method COCKATIEL uses NMF and sensitivity analysis to identify and rank concepts used by the model.
result COCKATIEL provides accurate and meaningful explanations without affecting model performance.