摘要:
to deal with underfitting增加feature的数目(通过feature的平方,立方项等增加feature或者增加其他的feature)减小λ的值to deal with overfitting找更多的训练样本减少feature的数目用regulation,增加λ的值θeval... 阅读全文
摘要:
Linear Regression with Multiple Variablesmean normalization:将不同特征的数据scale统一目的:为了使梯度下降法更快地找到全局最小值实现方法:把原始数据做变换,(Xraw-mean)/(XrawMax-XrawMin)feature sca... 阅读全文