|
|
#1
08-19-2012, 12:38 AM
|
|||
|
|||
Classification Error - HW6 Q2
The reference to classification error in question 2 means that we should just count the number of misclassified points rather than summing the squared differences to calculate the total error, even though linear regression was used to perform the classifications, right?
|
|
#2
08-19-2012, 12:48 AM
|
|||
|
|||
|
Re: Classification Error - HW6 Q2
Yes, I believe so. I had implemented the squared error and was getting weird results, but when I switched to just taking the sign of X*w' and comparing that with Y, counting up the mismatches and dividing by the number of examples, then I got results that fit with the choices given in the question.
|
|
#3
08-19-2012, 06:28 AM
|
|||
|
|||
|
Re: Classification Error - HW6 Q2
Yes, use classification error for Q2-Q6.
Last edited by the cyclist; 08-19-2012 at 06:28 AM. Reason: Typo correction
|
 |
| Thread Tools | |
| Display Modes | |
|
|
Linear Mode
