Ibrahima J. Ndiour

Ibrahima J. Ndiour, Nilesh A. Ahuja, Omesh Tickoo

This paper presents a fast, principled approach for detecting anomalous and out-of-distribution (OOD) samples in deep neural networks (DNN). We propose the application of linear statistical dimensionality reduction techniques on the semantic features produced by a DNN, in order to capture the low-dimensional subspace truly spanned by said features. We show that the "feature reconstruction error" (FRE), which is the $\ell_2$-norm of the difference between the original feature in the high-dimensional space and the pre-image of its low-dimensional reduced embedding, is highly effective for OOD and anomaly detection. To generalize to intermediate features produced at any given layer, we extend the methodology by applying nonlinear kernel-based methods. Experiments using standard image datasets and DNN architectures demonstrate that our method meets or exceeds best-in-class quality performance, but at a fraction of the computational and memory cost required by the state of the art. It can be trained and run very efficiently, even on a traditional CPU.

Ibrahima Ndiour, Nilesh Ahuja, Utku Genc et al.

This paper presents a fast and principled approach for solving the visual anomaly detection and segmentation problem. In this setup, we have access to only anomaly-free training data and want to detect and identify anomalies of an arbitrary nature on test data. We propose the application of linear statistical dimensionality reduction techniques on the intermediate features produced by a pretrained DNN on the training data, in order to capture the low-dimensional subspace truly spanned by said features. We show that the \emph{feature reconstruction error} (FRE), which is the $\ell_2$-norm of the difference between the original feature in the high-dimensional space and the pre-image of its low-dimensional reduced embedding, is extremely effective for anomaly detection. Further, using the same feature reconstruction error concept on intermediate convolutional layers, we derive FRE maps that provide pixel-level spatial localization of the anomalies in the image (i.e. segmentation). Experiments using standard anomaly detection datasets and DNN architectures demonstrate that our method matches or exceeds best-in-class quality performance, but at a fraction of the computational and memory cost required by the state of the art. It can be trained and run very efficiently, even on a traditional CPU.

Amanda Rios, Ibrahima Ndiour, Parual Datta et al.

In the field of continual learning, relying on so-called oracles for novelty detection is commonplace albeit unrealistic. This paper introduces CONCLAD ("COntinuous Novel CLAss Detector"), a comprehensive solution to the under-explored problem of continual novel class detection in post-deployment data. At each new task, our approach employs an iterative uncertainty estimation algorithm to differentiate between known and novel class(es) samples, and to further discriminate between the different novel classes themselves. Samples predicted to be from a novel class with high-confidence are automatically pseudo-labeled and used to update our model. Simultaneously, a tiny supervision budget is used to iteratively query ambiguous novel class predictions, which are also used during update. Evaluation across multiple datasets, ablations and experimental settings demonstrate our method's effectiveness at separating novel and old class samples continuously. We will release our code upon acceptance.

Amanda Rios, Ibrahima Ndiour, Parual Datta et al.

AI deployed in many real-world use cases should be capable of adapting to novelties encountered after deployment. Here, we consider a challenging, under-explored and realistic continual adaptation problem: a deployed AI agent is continuously provided with unlabeled data that may contain not only unseen samples of known classes but also samples from novel (unknown) classes. In such a challenging setting, it has only a tiny labeling budget to query the most informative samples to help it continuously learn. We present a comprehensive solution to this complex problem with our model "CUAL" (Continual Uncertainty-aware Active Learner). CUAL leverages an uncertainty estimation algorithm to prioritize active labeling of ambiguous (uncertain) predicted novel class samples while also simultaneously pseudo-labeling the most certain predictions of each class. Evaluations across multiple datasets, ablations, settings and backbones (e.g. ViT foundation model) demonstrate our method's effectiveness. We will release our code upon acceptance.

Amanda S. Rios, Ibrahima J. Ndiour, Parual Datta et al.

AI deployed in the real-world should be capable of autonomously adapting to novelties encountered after deployment. Yet, in the field of continual learning, the reliance on novelty and labeling oracles is commonplace albeit unrealistic. This paper addresses a challenging and under-explored problem: a deployed AI agent that continuously encounters unlabeled data - which may include both unseen samples of known classes and samples from novel (unknown) classes - and must adapt to it continuously. To tackle this challenge, we propose our method COUQ "Continual Open-world Uncertainty Quantification", an iterative uncertainty estimation algorithm tailored for learning in generalized continual open-world multi-class settings. We rigorously apply and evaluate COUQ on key sub-tasks in the Continual Open-World: continual novelty detection, uncertainty guided active learning, and uncertainty guided pseudo-labeling for semi-supervised CL. We demonstrate the effectiveness of our method across multiple datasets, ablations, backbones and performance superior to state-of-the-art.

Ergin Utku Genc, Nilesh Ahuja, Ibrahima J Ndiour et al.

This brief sketches initial progress towards a unified energy-based solution for the semi-supervised visual anomaly detection and localization problem. In this setup, we have access to only anomaly-free training data and want to detect and identify anomalies of an arbitrary nature on test data. We employ the density estimates from the energy-based model (EBM) as normalcy scores that can be used to discriminate normal images from anomalous ones. Further, we back-propagate the gradients of the energy score with respect to the image in order to generate a gradient map that provides pixel-level spatial localization of the anomalies in the image. In addition to the spatial localization, we show that simple processing of the gradient map can also provide alternative normalcy scores that either match or surpass the detection performance obtained with the energy value. To quantitatively validate the performance of the proposed method, we conduct experiments on the MVTec industrial dataset. Though still preliminary, our results are very promising and reveal the potential of EBMs for simultaneously detecting and localizing unforeseen anomalies in images.

Ibrahima Ndiour, Nilesh Ahuja, Omesh Tickoo

This paper presents a principled approach for detecting out-of-distribution (OOD) samples in deep neural networks (DNN). Modeling probability distributions on deep features has recently emerged as an effective, yet computationally cheap method to detect OOD samples in DNN. However, the features produced by a DNN at any given layer do not fully occupy the corresponding high-dimensional feature space. We apply linear statistical dimensionality reduction techniques and nonlinear manifold-learning techniques on the high-dimensional features in order to capture the true subspace spanned by the features. We hypothesize that such lower-dimensional feature embeddings can mitigate the curse of dimensionality, and enhance any feature-based method for more efficient and effective performance. In the context of uncertainty estimation and OOD, we show that the log-likelihood score obtained from the distributions learnt on this lower-dimensional subspace is more discriminative for OOD detection. We also show that the feature reconstruction error, which is the $L_2$-norm of the difference between the original feature and the pre-image of its embedding, is highly effective for OOD detection and in some cases superior to the log-likelihood scores. The benefits of our approach are demonstrated on image features by detecting OOD images, using popular DNN architectures on commonly used image datasets such as CIFAR10, CIFAR100, and SVHN.

Mahesh Subedar, Nilesh Ahuja, Ranganath Krishnan et al.

Data poisoning attacks compromise the integrity of machine-learning models by introducing malicious training samples to influence the results during test time. In this work, we investigate backdoor data poisoning attack on deep neural networks (DNNs) by inserting a backdoor pattern in the training images. The resulting attack will misclassify poisoned test samples while maintaining high accuracies for the clean test-set. We present two approaches for detection of such poisoned samples by quantifying the uncertainty estimates associated with the trained models. In the first approach, we model the outputs of the various layers (deep features) with parametric probability distributions learnt from the clean held-out dataset. At inference, the likelihoods of deep features w.r.t these distributions are calculated to derive uncertainty estimates. In the second approach, we use Bayesian deep neural networks trained with mean-field variational inference to estimate model uncertainty associated with the predictions. The uncertainty estimates from these methods are used to discriminate clean from the poisoned samples.

Nilesh A. Ahuja, Ibrahima Ndiour, Trushant Kalyanpur et al.

We present a principled approach for detecting out-of-distribution (OOD) and adversarial samples in deep neural networks. Our approach consists in modeling the outputs of the various layers (deep features) with parametric probability distributions once training is completed. At inference, the likelihoods of the deep features w.r.t the previously learnt distributions are calculated and used to derive uncertainty estimates that can discriminate in-distribution samples from OOD samples. We explore the use of two classes of multivariate distributions for modeling the deep features - Gaussian and Gaussian mixture - and study the trade-off between accuracy and computational complexity. We demonstrate benefits of our approach on image features by detecting OOD images and adversarially-generated images, using popular DNN architectures on MNIST and CIFAR10 datasets. We show that more precise modeling of the feature distributions result in significantly improved detection of OOD and adversarial samples; up to 12 percentage points in AUPR and AUROC metrics. We further show that our approach remains extremely effective when applied to video data and associated spatio-temporal features by detecting adversarial samples on activity classification tasks using UCF101 dataset, and the C3D network. To our knowledge, our methodology is the first one reported for reliably detecting white-box adversarial framing, a state-of-the-art adversarial attack for video classifiers.