2011/05/01 by Gang Niu, Bo Dai, Niu, Gang +5 · 2 citations
Computer Science · Mathematics · #Algorithm #Artificial Intelligence (cs.AI) #Artificial intelligence #Artificial neural network #Computer science #Entropy (arrow of time) #FOS: Computer and information sciences #Face and Expression Recognition #Gaussian Processes and Bayesian Inference #Machine Learning (stat.ML) #Machine learning #Mahalanobis distance #Mathematical optimization #Mathematics #Metric (unit) #Neural Networks and Applications #Parameterized complexity #Regularization (linguistics) #Semi-supervised learning #Supervised learning #cs.AI #stat.ML
paper · pdf · doi:10.48550/arxiv.1105.0167
published in arXiv (Cornell University) (Cornell University) · The same paper has been submitted to arXiv by ICML 2012. See http://arxiv.org/abs/1206.4614
openalex publication_date 2011/05/01 · arxiv created 2012/11/15 · arxiv updated 2015/03/18 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05
We propose a general information-theoretic approach called Seraph (SEmi-supervised metRic leArning Paradigm with Hyper-sparsity) for metric learning that does not rely upon the manifold assumption. Given the probability parameterized by a Mahalanobis distance, we maximize the entropy of that probability on labeled data and minimize it on unlabeled data following entropy regularization, which allows the supervised and unsupervised parts to be integrated in a natural and meaningful way. Furthermore, Seraph is regularized by encouraging a low-rank projection induced from the metric. The optimization of Seraph is solved efficiently and stably by an EM-like scheme with the analytical E-Step and convex M-Step. Experiments demonstrate that Seraph compares favorably with many well-known global and local metric learning methods.