LogLoss log类型损失函数的优势可以将连乘转为求和,由于是单调函数,不会改变原结果,并且还很方面求最优,因此log类型的损失函数函数也非常常用,比较著名的一种就是交叉熵(cross entropy),也就是logistic regression用的损失函数

L(Y,P(YX))=logP(YX) L(Y, P(Y|X)) = -logP(Y|X)

logistic regression Loss J(θ)=1mi=1m(y(i)log hθ(x(i))+(1y(i))log (1hθ(x(i)))) J(\theta) =-\frac{1}{m}\sum_{i=1}^m(y^{(i)} log\ h_\theta(x^{(i)}) + (1-y^{(i)})log\ (1-h_\theta(x^{(i)})))

应用:

  1. 逻辑回归 logistic regression

ref: https://people.eecs.berkeley.edu/~russell/classes/cs194/f11/lectures/CS194%20Fall%202011%20Lecture%2006.pdf

results matching ""

    No results matching ""