Find the following information about this article:
Title: Large scale metric learning from equivalence constraints
Author: Kostinger, Hirzer, Wohlhart, Roth, Bischof
Journal/Conference: CVPR
Year: 2012
![Graphical Abstract 1](images/D8_Graphical Abstract.jpg) ![Graphical Abstract 2](images/D9_Graphical Abstract.jpg)
- Different distance measure ?
- Mahalanobis distance?
- The topic indicates learning which correspond to what ?
- What are the main difference and similiarity between prpopsed method and PCA and LDA?
- What can be achieved from the distance?
- Regularization to avoid overfitting
- Maxliklihood --> Minimum distance
- How to minimize (Do we need optimization method ? Or we can solve it without them in much easier way using this method?)
Definition of metric learning
- Definition of Mahalanobis distance
- Specific case of Euclidean distance
- What Mahalahanobis and the covariance matrice will imply
![What is interesting in this paper](images/D2_Motivation_Minimization M_sans complex Alg.jpg)
Check the following notebook for an entire description of LMNN. ![missing image](images/D3_state of the art.jpg)
![missing image](images/D4_state of the art.jpg) Entropy driven optimisation.
![KISS formulation](images/D5_KISS_Equations and Formulations.jpg) ![KISS formulation](images/D6_KISS_Equations and_Formulations.jpg)
![missingimage](images/D7_Supervised Learning_Where to apply.jpg)