Learning a similarity metric discriminatively, with application to face verification

By Sumit Chopra, Raia Hadsell, Yann LeCun. 2005

4,573 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning, Signal Processing

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

In Signal Processing

Citations in the corpus

Dense Passage Retrieval for Open-Domain Question Answering (2020, methodology): “To ensure that the embedding can be pre-computed and indexed offline, the basic architecture follows the Siamese neural network design (Bromley et al., 1994; Chopra et al., 2005).”
Learning Dense Representations for Entity Retrieval (2019, background): “Dual encoder architectures have a long history (Bromley et al., 1994; Chopra et al., 2005; Yih et al., 2011), including for retrieval (Gillick et al., 2018).”
Representation Learning with Contrastive Predictive Coding (2018, methodology): “For example, the techniques proposed by [21, 22, 23] were based on triplet losses using a max-margin approach to separate positive from negative examples.”
Dual Contradistinctive Generative Autoencoder (2020, neutral): “Contrastive learning can be connected to metric learning [4, 9, 5].”
Imitating Interactive Intelligence (2020, methodology): “The Language Matching (LM) auxiliary task was partially inspired by developments in contrastive self-supervised learning (Chopra et al., 2005; Gutmann and Hyvärinen, 2010; van den Oord et al., 2018; Hénaff et al., 2019).”
Unsupervised Pretraining Transfers Well Across Languages (2020, background): “for speech and we focus on those based on contrastive learning [7, 14, 15].”
Meta-learning with differentiable closed-form solvers (2018, background): “In a Siamese network [33, 34, 18], Λ is a nearest neighbor classifier, employing inner-products or various distance functions.”
Learning to See by Moving (2015, methodology): “In contrast, our method uses egomotion as super-vision by predicting the transformation between two images using a siamese-like network model [11].”
Autoencoding beyond pixels using a learned similarity metric (2015, methodology): “Neural networks have been applied to metric learning in form of the Siamese architecture (Bromley et al., 1993; Chopra et al., 2005).”
Order-Embeddings of Images and Language (2015, background): “One line of work, exempliﬁed by Chopra et al. (2005) and ﬁrst applied to the caption-image relationship by Socher et al. (2014), requires the mapping to be distance-preserving: semantically similar objects are mapped to points that are nearby in the embedding space.”
Unsupervised Learning of Visual Representations Using Videos (2015, background): “Finally, our work is also related to metric learning via deep networks [51, 31, 4, 17, 15, 22, 54].”
Learning visual groups from co-occurrences in space and time (2015, methodology): “To model w ( A, B ) we use a Convolutional Neural Net (CNN) with a Siamese-style architecture (Figure 2, Chopra et al. (2005)), which we implement in Caffe (Jia et al. (2014)).”
Learning Visual Clothing Style with Heterogeneous Dyadic Co-Occurrences (2015, methodology): “Then, we show how to train a Siamese CNN to learn a feature transformation from the image space into the latent style space.”
Learning Spread-Out Local Feature Descriptors (2017, background): “Contrastive loss was originally proposed in [4] with ε = 0.”
Recent advances in convolutional neural networks (2015, background): “3) Contrastive loss: Contrastive loss is commonly used to train Siamese network [47] which is a weakly-supervised”