ftrl_trainer : trainPair函数，一对sample，为什么相同名称的 feature 要覆盖掉呢？

[YELLOWH]

for(unordered_map<string, double>::iterator iter = sample1.x.begin(); iter != sample1.x.end(); ++iter) { const string& index = iter->first; if(theta.find(index) == theta.end()) { theta[index] = pModel->getOrInitModelUnit(index); } }

for(unordered_map<string, double>::iterator iter = sample2.x.begin(); iter != sample2.x.end(); ++iter) { const string& index = iter->first; if(theta.find(index) == theta.end()) { theta[index] = pModel->getOrInitModelUnit(index); } }

看了楼主博客的公式推导，还是不太明白，望楼主看到的话可以解答下，谢谢啦