Hello,
I am just wondering in part b) of the question 30 if we should not write argmax y in {-1,1} instead of {0,1}? Because indeed, the output of S(a) is sign() so it's either -1 or 1 no?
Hi, should 0 also be included in y so that \(y\in\{-1,0,-1\}\) due to the definition of sign() function. In this question that \(\frac{1}{1+e^{-a}} - \frac{1}{2} = 0\)
Q30 exam 2019
Hello,
I am just wondering in part b) of the question 30 if we should not write argmax y in {-1,1} instead of {0,1}? Because indeed, the output of S(a) is sign() so it's either -1 or 1 no?
Hi,
yes, you are right that the output is in {-1, 1} and thus the solution should have y \in {-1, 1}.
Note that for logistic regression seen in the class y was considered to be in {0, 1}.
2
up.
Hi,
yes, you are right that the output is in {-1, 1} and thus the solution should have y \in {-1, 1}.
Note that for logistic regression seen in the class y was considered to be in {0, 1}.
2
Hi, should 0 also be included in y so that \(y\in\{-1,0,-1\}\) due to the definition of sign() function. In this question that \(\frac{1}{1+e^{-a}} - \frac{1}{2} = 0\)
Add comment