XEnsemble improves DNN robustness against adversarial and out-of-distribution inputs.
problem Protecting DNN models from adversarial and out-of-distribution inputs.
method Diverse input denoising verifiers and disagreement-diversity ensemble learning.
result XEnsemble achieves high defense and detection success rates.
This paper examines how the choice of prior distribution affects likelihoods of out-of-distribution inputs in deep generative models.
problem Mismatch between prior and data distributions causes deep generative models to assign higher likelihoods to out-of-distribution inputs.
method Proposes using a mixture distribution as a prior to make likelihoods of out-of-distribution inputs more sensitive.
result A mixture prior lowers the out-of-distribution likelihood with respect to real image data sets.
Deep learning provides a powerful tool for machine perception when the observations resemble the training data. However, real-world robotic systems must react intelligently to their observations even in unexpected circumstances. This requires a system to reason about its own uncertainty given unfamiliar, out-of-distrib…
Proposes training neural networks to predict uncertainty for out-of-distribution inputs.
problem Poor uncertainty predictions for out-of-distribution inputs limit model robustness.
method Generates pseudo-inputs in low-density regions and trains a Bayesian framework.
result Yields robust and interpretable uncertainty predictions.
Bayesian autoencoder detects AI safety risk of out-of-distribution inputs.
problem Detecting unreliable predictions from AI models with different distributions.
method Probabilistic, unsupervised Bayesian variational autoencoder with posterior estimation.
result Effective detection of out-of-distribution inputs in both input and latent spaces.
Improved VAE model enhances uncertainty estimation for out-of-distribution samples.
problem VAEs assign higher likelihood to out-of-distribution inputs.
method INCPVAE integrates noise contrastive prior into VAEs for reliable uncertainty estimation.
result INCPVAE outperforms standard VAEs in uncertainty estimation for OOD inputs.
In-N-Out improves model robustness to out-of-distribution data.
problem Learning robust models with few in-distribution labeled examples.
method Pre-training with auxiliary information and self-training with pseudolabels.
result In-N-Out outperforms auxiliary inputs or outputs alone on both in-distribution and OOD error.
Recent work has shown that deep generative models can assign higher likelihood to out-of-distribution data sets than to their training data (Nalisnick et al., 2019; Choi et al., 2019). We posit that this phenomenon is caused by a mismatch between the model's typical set and its areas of high probability density. In-dis…
Generative classifier derived from any discriminative classifier rejects illegal inputs.
problem Detecting and rejecting illegal inputs like adversarial examples and out-of-distribution samples.
method SDIM-logit: learns generative classifier from logits of any discriminative classifier, imposing statistical constraints.
result SDIM-logit inherits performance of base classifier without loss and can reject illegal inputs.
p-DkNN uses deep representations to detect out-of-distribution data with statistical tests.
problem Lack of reliable confidence estimates in neural networks for safety-critical applications.
method Statistical testing of deep neural network's intermediate hidden representations.
result p-DkNN enables more accurate and reliable predictions by abstaining from incorrect predictions.
New method improves robustness of OOD detection models.
problem Detecting out-of-distribution inputs is critical for deep learning models.
method Proposes ALOE algorithm for robust training with adversarially crafted examples.
result ALOE substantially improves robustness of OOD detection on CIFAR-10 and CIFAR-100 datasets.
We present a new method for uncertainty estimation and out-of-distribution detection in neural networks with softmax output. We extend softmax layer with an additional constant input. The corresponding additional output is able to represent the uncertainty of the network. The proposed method requires neither additional…
DOODLER detects out-of-distribution inputs by reconstructing in-distribution data.
problem Detecting real-world out-of-distribution inputs for deep learning models.
method DOODLER uses a Variational Auto-Encoder to reconstruct in-distribution data and identifies failures as out-of-distribution.
result DOODLER outperforms other OOD detection methods under similar constraints.
Deep learning models are known to be overconfident in their predictions on out of distribution inputs. This is a challenge when a model is trained on a particular input dataset, but receives out of sample data when deployed in practice. Recently, there has been work on building classifiers that are robust to out of dis…
Modern neural networks are very powerful predictive models, but they are often incapable of recognizing when their predictions may be wrong. Closely related to this is the task of out-of-distribution detection, where a network must determine whether or not an input is outside of the set on which it is expected to safel…
Likelihood-based generative models are a promising resource to detect out-of-distribution (OOD) inputs which could compromise the robustness or reliability of a machine learning system. However, likelihoods derived from such models have been shown to be problematic for detecting certain types of inputs that significant…
FOOD detects out-of-distribution samples quickly without needing OOD data.
problem Detecting out-of-distribution samples efficiently in neural networks.
method Extended DNN classifier with Gaussian layer and log likelihood ratio test.
result FOOD achieves state-of-the-art performance and is fast and applicable.
MINs learn inverse mappings for high-dimensional optimization problems.
problem Data-driven optimization with high-dimensional inputs and valid subsets.
method Model Inversion Networks (MINs) learn an inverse mapping from scores to inputs.
result MINs can scale to high-dimensional input spaces and handle both offline and active data.
Deep learning models are known to be overconfident in their predictions on out of distribution inputs. There have been several pieces of work to address this issue, including a number of approaches for building Bayesian neural networks, as well as closely related work on detection of out of distribution samples. Recent…
Single deep model detects out-of-distribution data with single forward pass.
problem Detecting out-of-distribution data points in neural networks.
method Deterministic uncertainty quantification (DUQ) using gradient penalty for reliable detection.
result Single model outperforms or matches ensemble methods in out-of-distribution detection.
Real-time detection of out-of-distribution data in CPS control systems.
problem Detecting out-of-distribution data in CPS control systems for safety.
method Inductive conformal prediction and anomaly detection using variational autoencoders and deep support vector data description.
result Efficient real-time detection with low false alarm rates and comparable execution time.
Deep neural networks achieve superior performance in challenging tasks such as image classification. However, deep classifiers tend to incorrectly classify out-of-distribution (OOD) inputs, which are inputs that do not belong to the classifier training distribution. Several approaches have been proposed to detect OOD i…
Survey connects three DNN generalization fields for better inference.
problem Assessing if DNNs generalize correctly at inference time.
method Investigates predictive uncertainty, out-of-distribution, and adversarial example detection.
result Connects three fields for better DNN inference.
New method detects uncertainty in neural networks for out-of-distribution detection.
problem Detecting out-of-distribution inputs to ensure model reliability.
method Predictive topological uncertainty (pTU) based on persistent homology.
result pTU provides a statistical framework for OOD detection.
This paper examines how optimization methods affect the reliability of detecting inputs outside a model's training distribution.
problem The unreliability of deep neural networks on out-of-distribution inputs.
method Analysis of optimization methods' impact on OOD detection approaches.
result Optimization methods significantly influence the robustness of OOD detection approaches.
Bayesian autoencoders improve OOD detection by addressing Bernoulli likelihood issues.
problem Out-of-distribution (OOD) detection fails with Bernoulli likelihood for certain datasets.
method Proposes Bayesian autoencoders and alternative likelihood models to fix the issue.
result Bayesian autoencoders and alternative likelihood models improve OOD detection accuracy.
Modern deep artificial neural networks have achieved great success in the domain of computer vision and beyond. However, their application to many real-world tasks is undermined by certain limitations, such as overconfident uncertainty estimates on out-of-distribution data or performance deterioration under data distri…
Work proposes a new framework to improve uncertainty estimation in deep Bayesian models.
problem Traditional training procedures underestimate uncertainty in NLMs, leading to unreliable predictions.
method Introduces a novel training framework that captures useful predictive uncertainties for out-of-distribution inputs.
result Demonstrates that traditional methods for NLMs significantly underestimate uncertainty and propose a new framework to address this issue.
We introduce a measure to quantify ambiguity in deep learning models, improving their reliability.
problem Deep learning models make mistakes on seemingly trivial cases and fail in recognizing what they don't know.
method We define ambiguity based on decision boundaries and convex hulls in feature space, developing a theoretical framework to identify unknowns.
result A single ambiguity measure can detect a significant portion of model mistakes, including adversarial and out-of-distribution inputs.
A large body of recent work has investigated the phenomenon of evasion attacks using adversarial examples for deep learning systems, where the addition of norm-bounded perturbations to the test inputs leads to incorrect output classification. Previous work has investigated this phenomenon in closed-world systems where …
A new framework detects anomalous inputs to DNNs.
problem Detecting adversarial and out-of-distribution inputs in DNNs.
method Unsupervised meta-algorithm with configurable components based on statistical testing.
result The proposed method effectively detects adversarial and OOD inputs.
TIE framework detects out-of-distribution samples and estimates uncertainty without external datasets.
problem Detecting and estimating uncertainty for out-of-distribution samples in neural networks.
method TIE framework extends a classifier to an (n+1)-class model, iteratively refining through training, inversion, and exclusion.
result Unified and interpretable framework for robust anomaly detection and calibrated uncertainty estimation.
ACNML method improves uncertainty estimation for deep networks.
problem Uncertainty estimation and calibration for deep neural networks under distribution shift.
method Approximate Bayesian inference to approximate CNML distribution.
result ACNML compares favorably to prior techniques for uncertainty estimation.
We consider the problem of detecting out-of-distribution images in neural networks. We propose ODIN, a simple and effective method that does not require any change to a pre-trained neural network. Our method is based on the observation that using temperature scaling and adding small perturbations to the input can separ…
Δ-UQ uses anchoring to estimate uncertainty in models.
problem Estimating uncertainty in predictive models.
method Anchoring input into a tuple for uncertainty estimation.
result Δ-UQ outperforms baselines in various use-cases.
Proposes a new criterion for reliable uncertainty estimation in deep neural networks.
problem Inability of existing approaches to provide reliable uncertainty estimates for deep neural networks.
method Develops a density uncertainty layer architecture that satisfies the proposed criterion.
result Density uncertainty layers provide more reliable uncertainty estimates and robust out-of-distribution detection.
Counterfactual data augmentations may not ensure OOD robustness if performed by a context-guessing machine.
problem Deep learning models lack out-of-distribution robustness due to reliance on spurious features.
method Theoretical analysis and demonstration of counterfactual data augmentations performed by a context-guessing machine.
result Counterfactual data augmentations by a context-guessing machine do not lead to robust OOD classifiers.
DAB enriches deep networks with uncertainty estimates using a codebook of training inputs.
problem Lack of uncertainty estimation in deep neural networks.
method Distance Aware Bottleneck (DAB) method that learns a codebook of training inputs.
result DAB achieves better OOD detection and misclassification prediction than prior methods.
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.
GIM uses neural networks with Gaussian distributions to detect out-of-distribution data.
problem Incorrect classification of out-of-distribution data by neural networks.
method GIM is a hybrid classifier based on neural networks with a new loss function that imposes Gaussian distributions on each class.
result GIM achieves state-of-the-art results on image recognition and sentiment analysis datasets.
Framework extends neural operators to handle functions outside training set.
problem Robust handling of functions beyond the training set.
method Kernel approximation techniques and Reproducing Kernel Hilbert Spaces (RKHSs) theory.
result Theoretical framework and empirical validation for reliable function extension.
Paper proposes contrastive training to improve OOD detection without needing explicit OOD examples.
problem Improving reliable detection of out-of-distribution inputs for machine learning systems.
method Contrastive training approach that doesn't require explicit OOD examples, using CLP score.
result Contrastive training significantly improves OOD detection performance on benchmarks, especially in near OOD classes.
Bayesian neural networks fail at out-of-distribution detection, revealing fundamental issues.
problem Out-of-distribution detection with Bayesian neural networks.
method Study of Bayesian inference with function space priors and comparison to Gaussian processes.
result Bayesian inference with function space priors does not lead to good OOD detection.
There is an increasingly apparent need for validating the classifications made by deep learning systems in safety-critical applications like autonomous vehicle systems. A number of recent papers have proposed methods for detecting anomalous image data that appear different from known inlier data samples, including reco…
Paper aims to ensure reliable detection of out-of-distribution data with certifiable worst-case guarantees.
problem Deep neural networks are overconfident with OOD inputs, posing safety risks.
method Enforces low confidence and bounds in an l∞-ball around OOD points using interval bound propagation (IBP). result Certifiable worst-case guarantees for OOD detection are possible without significant loss in accuracy.
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.
Improves regression models' performance on covariate shift.
problem Out-of-distribution generalization for regression.
method Spectrally adapting the weights of a pre-trained neural regression model.
result Spectral adaptation improves out-of-distribution performance.
Ensemble approaches for uncertainty estimation have recently been applied to the tasks of misclassification detection, out-of-distribution input detection and adversarial attack detection. Prior Networks have been proposed as an approach to efficiently \emph{emulate} an ensemble of models for classification by paramete…