Softmax Regression
From Ufldl
(→Cost Function) |
(→Properties of softmax regression parameterization) |
||
Line 185: | Line 185: | ||
Softmax regression has an unusual property that it has a "redundant" set of parameters. To explain what this means, | Softmax regression has an unusual property that it has a "redundant" set of parameters. To explain what this means, | ||
suppose we take each of our parameter vectors <math>\theta_j</math>, and subtract some fixed vector <math>\psi</math> | suppose we take each of our parameter vectors <math>\theta_j</math>, and subtract some fixed vector <math>\psi</math> | ||
- | from it, so that <math>\ | + | from it, so that <math>\theta_1</math> is now replaced with <math>\theta_1 - \psi</math>, |
+ | <math>\theta_2</math> is replaced with <math>\theta_2 - \psi</math>, and so on. Our hypothesis | ||
now estimates the class label probabilities as | now estimates the class label probabilities as | ||