Learning Discriminative Reconstructions for Unsupervised Outlier Removal

By Yan Xia, Xudong Cao, Fang Wen, Gang Hua. IEEE International Conference on Computer Vision, 2015

309 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/1c06870e1ecc63e120e45a2283ca4b72c153e867

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

In Computer Vision and Pattern Recognition

Citations in the corpus

CleanNet: Transfer Learning for Scalable Image Classifier Training with Label Noise (2017, methodology): “One-Class SVM [24], UOCL [17], and DRAE [34]).”
Image Classification with Deep Learning in the Presence of Noisy Labels: A Survey (2019, methodology): “[149] uses reconstruction error of autoencoder to discriminate noisy data from clean data, arguing that noisy data tend to have bigger reconstruction error.”
Autoencoders for Unsupervised Anomaly Segmentation in Brain MR Images: A Comparative Study (2020, methodology): “[10] made an initial attempt towards this direction by using a percentile-based heuristic on the training data to mask out potential outliers during training, and with so called discriminative reconstruction autoencoders [24] an interesting concept has recently been proposed in the Computer Vision f”
OCGAN: One-Class Novelty Detection Using GANs With Constrained Latent Representations (2019, background): “In contrast, anomaly detection (also known as outlier detection) is an unsupervised learning task [30].”
Machine Learning Methods for Histopathological Image Analysis (2017, background): “) [114] have been proposed for outlier detection in other domains, but they are not yet applied in the histopathological image analysis.”
OpenGAN: Open-Set Recognition via Open Data Generation (2021, unknown): “As a result, previous GAN-based methods focus on 1) generating fake open-set data to augment the training set, and 2) relying on the reconstruction error for open-set recognition [53, 47, 44, 1, 13].”
Generative Probabilistic Novelty Detection with Adversarial Autoencoders (2018, methodology): “ditsabilitytodescribethetrainingdata. Thenit transformedtheimplicitdatadescriptionofnormaldatatoa noveltyscore. [10]trainedGANs using optical ﬂow images to learn a representation of scenes in videos. [7] minimized the reconstruction error of an autoencoderto remove outliers from noisy data, and by u”
Generative-Discriminative Feature Representations for Open-Set Recognition (2020, background): “Conditioned AE [28] is a modified version of Vanilla AE, where a fully connected layer classifier is connected to the latent space.”
HRN: A Holistic Approach to One Class Learning (2020, background): “More recently, deep learning models have been proposed for the same purpose [68, 8], which mainly learn a good latent representation of the given class of data using various auto-encoders [1, 14, 65, 71, 83, 69] or GAN [27] style adversarial training [74, 72, 15, 64].”
Deep Anomaly Detection Using Geometric Transformations (2018, methodology): “[32] use a convolutional autoencoder with a regularizing term that encourages outlier samples to have a large reconstruction error.”
Generalized Out-of-Distribution Detection: A Survey (2021, unknown): “Remark: Nuance between AD and ND Apart from the special interest in semantics, some literature [56], [57] also point out that ND is supposed to be fully unsupervised, while AD might have some abnormal training samples.”
Adversarially Learned One-Class Classifier for Novelty Detection (2018, methodology): “Such problems are especially of great interest in computer vision studies, as they are closely related to outlier detection [47, 52], image denoising [8], anomaly detection in images [10, 23] and videos [39].”
Classification-Based Anomaly Detection for General Data (2020, methodology): “Most recently, neural networks were used (Sakurada & Yairi, 2014; Xia et al., 2015) for learning deep basis functions for reconstruction.”