DANCE improves saliency maps by adding subtle input variations.
problem Poor performance of saliency methods in saturated gradients, adversarial perturbations, and inter-feature dependence.
method Two-step procedure: 1) Perturbation mechanism, 2) Aggregation of saliency maps.
result DANCE saliency method outperforms existing methods qualitatively and quantitatively.
Saliency methods often misattribute predictions due to input transformations.
problem Saliency methods lack reliability when explanations are sensitive to non-contributing factors.
method Used a simple pre-processing step to demonstrate that transformations with no effect on the model can cause misleading attributions.
result Saliency methods that do not satisfy input invariance (mirror model sensitivity to input transformations) result in misleading attributions.
Bayesian optimisation generates saliency maps for black-box models.
problem Generating saliency maps for models without access to parameters.
method Bayesian optimisation sampling method to find global salient regions.
result Approach outperforms grid-based methods and performs similarly to gradient-based methods.
The paper benchmarks saliency methods for time series predictions.
problem Benchmarking saliency methods for time series predictions.
method Compared saliency-based interpretability methods across diverse neural architectures.
result Saliency methods fail to reliably identify feature importance over time in time series data.
Deep learning predicts chess player attention using eye-tracking and game data.
problem Predicting chess player attention during game.
method Convolutional neural networks with skip-layer architecture and autoencoder.
result Deep features from natural images help in predicting saliency maps for chess.
Paper proposes a method to detect adversarial examples using saliency.
problem Detecting adversarial perturbations in machine learning models.
method Trains a binary classifier with origin data and saliency data.
result Shows good performance in detecting adversarial perturbations.
The paper evaluates saliency methods for image predictions, finding some inadequate.
problem The reliability of saliency methods in explaining model predictions.
method Proposed a methodology to evaluate saliency methods based on their independence from model and data.
result Some saliency methods are independent of both model and data, making them inadequate for certain tasks.
Visualizes ConvNets without confounding effects.
problem Misinterpretation of saliency maps due to confounding variables.
method Univariate statistical tests and partial back-propagation to remove confounding effects.
result Visualization of confounder-free saliency maps.
Faster gaze prediction with less parameters.
problem Overparameterized networks for gaze prediction.
method Fisher pruning combined with knowledge distillation.
result 10x speedup for fixation prediction.
Predicts failure of autonomous vehicle steering control models.
problem Evaluating and predicting failure of machine learning models in safety-critical applications.
method Trains a student model to predict the main model's error based on saliency maps.
result Preliminary results show the failure predictor model works on autonomous vehicle steering control systems.
New method improves interpretability of fMRI decoding models.
problem Uninterpretable deep neural networks in fMRI decoding.
method Adversarial training to make DNNs robust to noise and improved saliency map methods.
result Saliency maps from adversarial-trained DNNs are more interpretable than those from other methods.
Proposes a method to quantify the reliability of salient regions in deep learning models using p-values.
problem Difficulty in assessing the reliability of saliency maps generated by deep learning models.
method Proposes a selective inference framework to quantify the reliability of salient regions as selected hypotheses by deep learning models.
result The method can provably control the probability of false positive detections of salient regions.
New method for visualizing high-level concepts in generative models.
problem Challenges in evaluating and visualizing concepts in generative models.
method Introduces a method to compute concept saliency maps for latent representations of known or novel high-level concepts.
result Concept saliency maps highlight input features important for high-level concepts.
Input-cell attention improves RNN saliency detection over time.
problem Vanishing saliency in RNNs makes them unreliable for detecting important features at arbitrary time intervals.
method Proposes input-cell attention, a novel RNN cell structure that uses a fixed-size matrix embedding to attend to different inputs from current or previous time steps.
result Input-cell attention produces a saliency map that can detect important features regardless of their occurrence in time.
Method uses ANN to estimate incentive salience from large behavioral data.
problem Estimating incentive salience in naturalistic settings.
method Artificial Neural Networks (ANNs) for latent state approximation.
result ANNs produce better representations for predicting future behaviour.
A method to combine saliency metrics for better CNN pruning decisions.
problem Improving CNN pruning decisions by combining multiple saliency metrics.
method Proposes a method to compose different saliency metrics for better CNN pruning decisions.
result The composition of saliencies avoids many poor pruning choices identified by individual saliencies.
The study evaluates saliency metrics for image classifier outputs, finding inconsistencies and unreliability.
problem Inconsistencies and unreliability in saliency metrics for evaluating pixel relevance.
method Investigated existing saliency metrics, calculated and compared their consistency, and applied psychometric testing methods.
result Saliency metrics can be statistically unreliable and inconsistent, affecting comparative rankings.
Develops methods to show model confidence and feature importance in medical imaging.
problem Ensuring safety and understanding model confidence in medical applications.
method Creates a pipeline to visualize uncertainty and saliency maps for deep neural networks.
result Demonstrates how deep neural networks can be made more transparent and safe for medical applications.
New hypothesis and method reduce noise in saliency maps.
problem Noisy saliency maps from deep neural networks.
method Proposed a new hypothesis and Rectified Gradient method.
result Rectified Gradient effectively reduces saliency map noise.
Bayesian explanations are more resilient to adversarial attacks than deterministic ones.
problem Stability of saliency-based explanations under adversarial attacks in Neural Networks.
method Empirical and theoretical analysis of Bayesian vs deterministic Neural Networks.
result Bayesian explanations are more stable under adversarial perturbations and direct attacks.
Convolutional-deconvolution networks can be adopted to perform end-to-end saliency detection. But, they do not work well with objects of multiple scales. To overcome such a limitation, in this work, we propose a recurrent attentional convolutional-deconvolution network (RACDNN). Using spatial transformer and recurrent …
Novel method combines neural network features with survival models for ICU infections.
problem Improving predictive models of ICU infections while maintaining interpretability.
method Semi-parametric approach combining low-resolution and high-resolution data.
result Improved predictive power with interpretability maintained.
A taxonomy of saliency metrics helps in pruning deep neural networks.
problem Difficulty in separating saliency metric effectiveness from pruning algorithms.
method Proposed a taxonomy based on four orthogonal components.
result Constructed metrics can outperform existing state-of-the-art metrics.
XRAI improves saliency maps for deep neural networks.
problem Improving understanding of deep neural networks through better saliency methods.
method XRAI is a region-based attribution method built upon integrated gradients, with new evaluation methods and sanity checks.
result XRAI produces better results than other saliency methods for common models and the ImageNet dataset.
Certifiable defense method improves robustness of deep learning interpretation.
problem Vulnerability of gradient-based saliency maps to adversarial attacks.
method Sparsified SmoothGrad method, extending certifiably robust smooth classifier bounds.
result Sparsified SmoothGrad method is certifiably robust against adversarial perturbations.
Study evaluates saliency maps on artificial data with different backgrounds.
problem Objective evaluation of saliency methods on artificial data with varying backgrounds.
method Developed a framework to generate artificial data with synthetic lesions and a known ground truth map, evaluated two data sets with different backgrounds (Perlin noise and 2D brain MRI slices).
result Heatmaps vary strongly between saliency methods and backgrounds.
New nonlinear saliency maps improve deep neural network interpretability.
problem Lack of understanding why and how deep neural networks make decisions.
method Developed novel nonlinear saliency maps to better interpret deep neural networks.
result Nonlinear saliency maps provide more specific drivers of classification on complex examples.
GGA improves untrustworthy prediction detection in neural networks without retraining.
problem Susceptibility of neural networks to untrustworthy predictions, especially adversarial attacks and out-of-distribution data.
method Geometric Gradient Analysis (GGA) analyzes the geometry of neural network loss landscapes based on saliency maps.
result GGA outperforms existing methods in detecting untrustworthy predictions, including adversarial and out-of-distribution data.
In this work we develop a fast saliency detection method that can be applied to any differentiable image classifier. We train a masking model to manipulate the scores of the classifier by masking salient parts of the input image. Our model generalises well to unseen images and requires a single forward pass to perform …
Study validates saliency maps of GNNs using selective inference.
problem Reliability of GNN saliency maps in graph-structured data.
method Statistical testing framework with selective inference to control Type I error rate.
result Valid p-values for salient subgraphs, ensuring meaningful information. Adversarial robust models have more interpretable saliency maps.
problem Understanding the relationship between adversarial robustness and saliency map interpretability.
method Quantifying the alignment between input images and saliency maps, hypothesizing and testing the relationship with linear and neural network models.
result The alignment between input images and saliency maps increases as the distance to the decision boundary grows, especially in linear models.
New method extracts high-quality saliency maps without classifier dependence.
problem Fixed classifier dependence limits saliency map extraction performance.
method Classifier-agnostic saliency map extraction.
result Sets new state of the art for localization tasks on ImageNet.
Study identifies and mitigates causes of image misclassifications in CNN models.
problem Improving model interpretability and accuracy in image classification.
method Trained six CNN architectures on CIFAR-10, used conditional confusion matrices and misclassification networks to identify morphological similarity and non-essential information interference as causes of misclassification. Developed a method to reduce misclassifications by erasing pixels within top 5% saliency map bounding boxes.
result Identified two causes of misclassification: morphological similarity and non-essential information interference, and developed a method to reduce misclassifications caused by the latter.
Contextual PDA improves explanation of image classifications for saturated models.
problem Difficulty in explaining decisions of saturated classifiers.
method Proposes Contextual PDA, a faster method for explaining image classifications.
result Contextual PDA outperforms PDA in explaining image classifications of state-of-the-art deep networks.
A new LDA variant improves multi-label classification performance.
problem Improving multi-label classification performance.
method Saliency-based weights redefine between-class and within-class scatter matrices for multi-label classification.
result The proposed method leads to performance improvements in various multi-label classification problems.
New approach makes adversarial examples less suspicious without changing perceptual salience.
problem Robustness of deep neural networks to unsuspicious adversarial examples.
method Splitting images into foreground and background, allowing larger perturbations in background while maintaining low cognitive salience.
result Dual-perturbation attacks are effective against classifiers robust to conventional attacks and adversarial training yields more robust classifiers.
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.
New saliency evaluations focus on completeness and soundness, improving explanations.
problem Current saliency evaluations focus on completeness but ignore soundness.
method Introduces new intrinsic evaluation metrics based on completeness and soundness.
result Simple saliency method matches or outperforms prior methods in new evaluations.
A simple fix helps saliency methods pass sanity checks.
problem Saliency methods fail simple sanity checks.
method Compute saliency maps for all labels, use competition to identify and remove less relevant pixels.
result CGI method is efficient, requires no additional training, and uses only the input and gradient.
Develops a new feature selection method using deep-learning saliency.
problem Lack of instance-level feature selection information.
method Saliency-based Feature Selection (SFS) method.
result SFS method provides instance-level feature selection information.
Puzzle Mix optimizes mixup by leveraging saliency and local statistics.
problem Overfitting and adversarial vulnerability of deep networks.
method Puzzle Mix alternates between multi-label and saliency-based objectives for optimal mixing masks.
result Puzzle Mix outperforms other mixup methods in generalization and adversarial robustness.
Proposes novel method for detecting novel scenarios in autonomous systems.
problem Detecting when a machine learning model makes a trustworthy prediction in dynamic, real-world situations.
method Leverages trained model's learned information and a new image similarity metric.
result Demonstrates the method's efficacy on real-world driving and indoor racing datasets.
1-Lipschitz neural networks produce clearer, more focused Saliency Maps for explainable AI.
problem Noisy and limited Saliency Maps from traditional neural networks.
method Dual loss of optimal transport problem for 1-Lipschitz neural networks.
result Saliency Maps from 1-Lipschitz networks are highly concentrated and less noisy, aligning with human explanations.
Enhances counterfactual explanations with more valid and informative saliency maps.
problem Lack of valid counterfactual explanations in existing models.
method Introduces a modified approach to CELS model by removing mask normalization.
result Demonstrates higher validity and more informative counterfactual explanations.
Develops deep neural networks for efficient salient object detection.
problem Efficiency and accuracy in detecting salient objects of various scales and semantic information.
method Hybrid contrast-oriented deep neural networks combining fully convolutional and segment-level spatial pooling streams, with an attentional module for fusion.
result Significantly outperforms state-of-the-art methods on six benchmark datasets.
Analyzes SmoothGrad and VarGrad methods for saliency maps.
problem Lack of rigorous theoretical interpretation of noise-adding methods.
method Analytically formalizes SmoothGrad and VarGrad techniques.
result SmoothGrad does not smooth the gradient of the score function.
New framework learns phoneme metric from perception data.
problem Learn metric for phoneme similarity from data.
method Learning algorithms to derive metric from behavioral data.
result Framework outperforms previous metrics in phoneme prediction.
A saliency detection method for ECT images segments cellular components without supervision.
problem Automatic segmentation of cellular components from ECT images is difficult due to structural complexity and imaging limits.
method Supervoxel over-segmentation, feature extraction, feature matrix decomposition, and computation of saliency.
result The method successfully labels most salient regions detected by a human observer and filters out background regions.