Even with the advent of more sophisticated, data-hungry methods, boosted
decision trees remain extraordinarily successful for fast rigid object
detection, achieving top accuracy on numerous datasets. While effective, most
boosted detectors use decision trees with orthogonal (single feature) splits,
and the topology of the resulting decision boundary may not
本研究提出了一种优化多元线性阈值函数作为决策树分裂函数以创建改进的 Random Forest 分类器的新算法，相较于标准的树归纳方法，该方法采用随机梯度下降优化每个节点的线性组合（斜）分裂函数，比之前的构建斜树的技术和单变量分裂函数的随机抽样和穷举搜索方法具有更好的性能，已在多类分类基准测试和 Labeled Faces in the Wild （LFW）数据集上进行了实验验证。