Towards good practice in large-scale learning for image classification

Perronnin, Florent, et al. "Towards good practice in large-scale learning for image classification." Computer Vision and Pattern Recognition (CVPR), 2012 IEEE Conference on. IEEE, 2012.

大規模的影像分類在現在大量的被運用，而現在的技術大部分是運用高維的image descripter在加上linear classifiers。會用linear classifiers的原因則是因為計算快速。

Notations:

其中前者指的是empirical risk
而後者指的是regularization penalty

SVM:

Binary One-Vs-Rest SVM (OVR):
assume only two classes







Multiclass SVM (MUL)





Ranking SVM (RNK)

Weighted Approximate Ranking SVM (WAR)

Data reweighting
當traning資料不平衡的時候，Data reweighting通常會有幫助











Optimization Algorithms

To handle large datasets. we employ Stochastic Gradient Descent (SGD)

Implementation details
Bias
Stopping criterion.
Regularization.
Step size.
Reweighting.

Result