Out-of-Distribution Detection using Multiple Semantic Label Representations

By Gabi Shalev, Yossi Adi, Joseph Keshet. 2018

93 citations recorded by the source corpus.

Deep Neural Networks are powerful models that attained remarkable results on a variety of tasks. These models are shown to be extremely efficient when training and test data are drawn from the same distribution. However, it is not clear how a network will act when it is fed with an out-of-distribution example. In this work, we consider the problem of out-of-distribution detection in neural networks. We propose to use multiple semantic dense representations instead of sparse representation as the target label. Specifically, we propose to use several word representations obtained from different corpora or architectures as target labels. We evaluated the proposed model on computer vision, and speech commands detection tasks and compared it to previous methods. Results suggest that our method compares favorably with previous work. Besides, we present the efficiency of our approach for detecting wrongly classified and adversarial examples.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shift (2018, methodology): “[44] employ model ensembling to further improve detection reliability.”
Does Your Model Know the Digit 6 Is Not a Cat? A Less Biased Evaluation of "Outlier" Detectors (2018, background): “tion to detect OOD samples. Further extensions rely on statistics of hidden representations [27, 31], construct classiﬁer ensembles with subsets of data [39], or perform regression on word embeddings [36]. There are also ideas that rely on GANs [7, 21, 26, 34] to detect anomalies or novelty in the d”
Confidence Calibration and Predictive Uncertainty Estimation for Deep Medical Image Segmentation (2019, methodology): “[40] used multiple semantic dense representations of the target”
Detecting semantic anomalies (2019, background): “Among existing works, the closest in terms of approaching the essence of our discussion about semantic properties being the discriminative factor of interest is [9], where the training set is augmented with semantically similar labels.”
A Less Biased Evaluation of Out-of-distribution Sample Detectors (2018, methodology): “Further extensions rely on statistics of hidden representations [25, 29], construct classiﬁer ensembles with subsets of data [37], or perform regression on word embeddings [34].”
Unsupervised Out-of-Distribution Detection by Maximum Classifier Discrepancy (2019, unknown): “While previous studies [9, 14, 16, 25, 26] only use labeled ID data to train the neural network under supervision, we also utilize unlabeled data in the training process.”
Generalized Out-of-Distribution Detection: A Survey (2021, unknown): “In [139], the sparse one-hot labels are replaced with several dense word embeddings from different NLP models, forming multiple regression heads for robust training.”
A Unifying Review of Deep and Shallow Anomaly Detection (2020, methodology): “Consequently, the OOD problem reduces to determining whether x lies outside regions of high density in P+, which is exactly the AD problem that we have described in this review.”
Enhancing the Robustness of Prior Network in Out-of-Distribution Detection (2018, methodology): “In Table 2, we mainly compare against Baseline [7], ODIN [14], Conﬁdence [3] and Semantic [22] under the VGG13 architecture.”
Contrastive Training for Improved Out-of-Distribution Detection (2020, background): “[33] to improve the intermediate feature map for OOD detection.”
Generalized ODIN: Detecting Out-of-Distribution Image Without Learning From Out-of-Distribution Data (2020, methodology): “Some previous works have done a similar attempt to learn without OoD data, such as [35], which uses word embeddings for extra supervision, and [25] which applies metric learning criteria.”