Information-theoretic metric learning

By Jason V. Davis, Brian Kulis, Prateek Jain, Suvrit Sra, Inderjit S. Dhillon. 2006

2,131 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistical and Nonlinear Physics

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 Statistical and Nonlinear Physics

Citations in the corpus

Deep Weighted Averaging Classifiers (2018, methodology): “This is closely related to a long line of past work on metric learning [3, 10, 23, 58, 60], but rather than trying to optimize a particular notion of distance (such as Mahalanobis distance), we make use of a fixed distance function, and allow the architecture of standard deep models to do the equiva”
Learning to cluster in order to Transfer across domains and tasks (2017, methodology): “Therefore we have shown results for K-means (MacQueen et al., 1967), LPNMF (Cai et al., 2009), LSC (Chen & Cai, 2011), ITML (Davis et al., 2007), SKKm (Anand et al., 2014), SKLR (Amid et al., 2016), SKMS (Anand et al., 2014), CSP (Wang et al., 2014) , and MPCK-means (Bilenko et al., 2004) as our…”
Petuum: A New Platform for Distributed Machine Learning on Big Data (2013, background): “Learning a proper distance metric [8, 29] is essential for many distance based data mining and machine learning algorithms, such as retrieval, k-means clustering and k-nearest neighbor (k-NN) classiﬁcation.”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, methodology): “nant regularized Distance Metric Learning (L-DML) and the Manifold regularized Distance Metric Learning (M-DML) for adapting face recognition systems. The Information-Theoretic Metric Learning (ITML) [60] was used by Saenko et al. in [185] to directly learn a distance metric across different domains”
A Survey on Metric Learning for Feature Vectors and Structured Data (2013, methodology): “…been shown to be kernelizable (see for instance Schultz and Joachims, 2003; Shalev-Shwartz et al., 2004; Hoi et al., 2006; Torresani and Lee, 2006; Davis et al., 2007) using specific arguments, but in general kernelizing a particular metric algorithm is not trivial: a new formulation of the…”
Cross Dataset Person Re-identification (2014, methodology): “Davis et al.[8] presented an approach called LTML to formulate the learning of Mahalanobis distance function the problem as that of minimizing the differential relative entropy between two multivariate Gaussians under constraints on the distance function.”
Learning a Mahalanobis distance metric for data clustering and classification (2008, background): “developed an information theory-based framework [39].”
Learning a Discriminative Null Space for Person Re-identification (2016, methodology): “Among them, the distance metric learning based methods are RPLM [15], MtMCML [27], Mid-level Filter [41], SCNCD [38], Similarity Learning [3], LADF [21], ITML [5], LMNN [35], KISSME [17], and MCML [9], whilst the others are discriminative subspace learning based methods including kCCA [24], MFA [36]”
Large-scale Classification of Fine-Art Paintings: Learning The Right Metric on The Right Feature (2015, background): “Davis et al [11] introduce the measure of LogDet divergence regularization between two matrices M, M (cid:48) (can be interpreted as metrics).”
Relaxed Pairwise Learned Metric for Person Re-identification (2012, background): “1 sec LMNN [16] 18 59 75 91 97 75 sec ITML [17] 14 52 71 90 98 16 sec LDML [18] 5 21 30 51 71 0.”
Semantic Autoencoder for Zero-Shot Learning (2017, methodology): “[32] KISSME [30] ITML [16] LMNN [61] MLCA [33] Test loss 0.”
Adapting Visual Category Models to New Domains (2010, methodology): “Furthermore, the learned kernel function may be computed over arbtirary points, and the method may be scaled for very large data sets; see [8,14] for details.”
Learning Locally-Adaptive Decision Functions for Person Verification (2013, methodology): “Figure 5(a) compares our proposed method with classic ML algorithms: LMNN [24] and ITML [5], using the same feature.”
Identifying Ambiguous Similarity Conditions via Semantic Matching (2022, background): “With the guidance of the comparison relationship between object pairs [6], triplets [39, 43], and higher-order statistics [17], visually similar instances are pulled together, and visually dissimilar ones are pushed away.”
Latent Coincidence Analysis: A Hidden Variable Model for Distance Metric Learning (2012, methodology): “Several previous methods have been proposed for this problem, including neighborhood component analysis (NCA) [4], large margin neighbor neighbor classiﬁcation (LMNN) [5], and information-theoretic metric learning (ITML) [6].”
Semi-supervised Kernel Metric Learning Using Relative Comparisons (2016, methodology): “Our algorithm also uses the log det divergence, and we extend the technique of Davis et al. [2007] to handle relative comparisons.”
Person Re-identification by Multi-Channel Parts-Based CNN with Improved Triplet Loss Function (2016, neutral): “3 −− −− ITML[5] 16.”
Graph Matching Networks for Learning the Similarity of Graph Structured Objects (2019, background): “…learning assumes that the data already lies in a vector space, and only a linear metric matrix is learned to properly measure the distance in this space to group similar examples together and dissimilar examples to be far apart (Xing et al., 2003; Weinberger & Saul, 2009; Davis et al., 2007).”