This package includes two approaches to multi-class cost-sensitive learning, that is, mcKLR and mckNN. The descr iption of mcKLR can be found in our TPAMI paper and CVPR'08 paper; the descr iption of mckNN can be found in our TPAMI paper. Both approa

多种多任务学习算法的工具箱 有代码有使用手册 matlab版-the toolbox of multi-task learning algorithms (matlab version)

Duchi and Singer [24] proposed a boosting method for multi-class classification problems by utilizing the structural sparsity of model parameters. They claimed that the method can be generalized for multi-task learning.

论文 Least Squares Loss for Multi-class (task) Learning via the sparse group Lasso penalty 的算法源代码，是sparse+l1+lq,1的稀疏属性选择算法。- Least Squares Loss for Multi-class (task) Learning via the sparse group Lasso penalty Problem

Fast R-CNN是在R-CNN的基础上进行的改进，大致框架是一致的。总体而言，Fast R-CNN相对于R-CNN而言，主要提出了三个改进策略： 1. 提出了RoIPooling，避免了对提取的region proposals进行缩放到224x224，然后经过pre-trained CNN进行检测的步骤，加速了整个网络的learning与inference过程，这个是巨大的改进，并且RoIPooling是可导的，因此使得整个网络可以实现end-to-end learning，这个可以认为是