Similarity-based Classification: Concepts and Algorithms

By Yihua Chen, Eric K. Garcia, Maya R. Gupta, Ali Rahimi, Luca Cazzanti. Journal of machine learning research, 2009

404 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/ac81da0dac77e231cbf07136a8186f53e818793e

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

Semi-Supervised Learning via Generalized Maximum Entropy (2010, methodology): “provide an overview of classification algorithms using similarities as features and compare the use of similarity features versus kernels for various kernel based discriminative algorithms (Chen et al., 2009).”
Proposal : Scalable , Active and Flexible Learning on Distributions (2015, neutral): “Chen et al. (2009) study this setting, presenting four methods to make K̂ psd: • Spectrum clip: Set any negative eigenvalues in the spectrum of K̂ to zero.”
Secure Kernel Machines against Evasion Attacks (2016, methodology): “This is a well-known technique in similarity-based classification, which amounts to learning the SVM on the squared kernel [10, 27].”
Sparse Coding and Dictionary Learning for Symmetric Positive Definite Matrices: A Kernel Approach (2012, methodology): “For values of σ outside of the above set, it is possible to convert a pseudo kernel into a true kernel, as discussed for example in [20].”
Large Scale Online Learning of Image Similarity Through Ranking (2009, methodology): “estimates of the positive sub-space are very noisy when only based on a few samples (see also Chen et al. 2009, Section 2.1). Indeed, accurate estimation of the negative subspace is known to be a hard problem, because small perturbations can turn a negative but small eigenvalue, into a small but pos”
Linear-Time Learning on Distributions with Approximate Kernel Embeddings (2015, background): “Typical approaches involve eigendecomposing the Gram matrix, which usually costs O(N(3)) computation and also presents challenges for traditional inductive learning, where the test points are not known at training time (Chen et al. 2009).”
Multi-Task Averaging (2011, methodology): “We report classification results for six different benchmark similarity datasets (Chen et al., 2009).”
Optimal Teaching for Limited-Capacity Human Learners (2014, background): “Exemplar models have proven successful in accounting for human learning performance [13, 14], are consistent with neural representations of acquired categories [15], and share strong theoretical connections with machine learning approaches [16, 17].”