一. 传统的距离度量学习方法：

监督方法

1.NIPS 2005(LMNN):Distance Metric Learning for Large Margin Nearest Neighbor Classification

1.1 AAAI 2017(LMNN的拓展):Parameter Free Large Margin Nearest Neighbor for Distance Metric Learning

2. Lecture Notes(FLD（LDA的一种)，必看）：Fisher Linear Discriminant Analysis(FLD)

2.1 NIPS 2003(MMC,解决了LDA样本少出现不可逆的问题)：Efficient and Robust Feature Extraction by Maximum Margin Criterion

2.2 CVPR 2007(ANMM，注意与LDA的关系以及与MMC,LMNN的区别联系):Feature Extraction by Maximizing the Average Neighborhood Margin

2.3 ICCV 2005(SNMMC，同样是解决LDA出现的四大问题，逐步的思想以及margin的设定很不错)：Face Recognition By Stepwise Nonparametric Margin Maximum Criterion

2.4 ECML 2004(MMDA,将分类器与数据降维特征提取结合,非常棒的发现):Margin Maximizing Discriminant Analysis

3.NIPS 2009(相似度学习（Similarity Learning），放弃了对称和半定的约束，但实验效果也不变差，建议看一下):An Online Algorithm for Large Scale Image Similarity Learning

无监督方法：

1.Lecture Notes(PCA，必看）:Principal Component Analysis(PCA)

二. 深度距离度量学习：

1.CVPR 2014: Discriminative Deep Metric Learning for Face Verification in the Wild

2.CVPR 2015(注意与1的区别与联系):Multi-Manifold Deep Metric Learning for Image Set Classification

3.CVPR 2016(最大化使用batch)Deep Metric Learning via Lifted Structured Feature Embedding

4.NIPS 2016(注意与3的区别)Improved Deep Metric Learning with Multi-class N-pair Loss Objective

三. 辅助论文学习（上述距离度量论文中涉及的内容）：

1.CVPR2016（近似近邻搜索）: FANNG: Fast Approximate Nearest Neighbour Graphs