misclassified error and Eucledian distance

[rsridharan]

Dear Professor Abu-Mostafa;



I am actively taking part in your online course. I have been regulary taking the homework assignments you have been posting.

I encountered some difficulty with the home work #6 , especially for problems 2, 3, 4,5 and 6.

I am not clear on how you define classification error, which is a dimesionless fraciton of the miss-calssified pointes, also having a dimension in Euclidean distance.

I ended up with incorrect answers.


Since a clarification of this point will be necessary, to proceed with problems of howme work #7, I would very much apprecate your response to this.


Kind Regards.

[yaser]

Originally Posted by rsridharan

I encountered some difficulty with the home work #6 , especially for problems 2, 3, 4,5 and 6.

I am not clear on how you define classification error, which is a dimesionless fraciton of the miss-calssified pointes, also having a dimension in Euclidean distance.

Hi,

I think I see the source of confusion. The classification error is correctly how you define it, and indeed it is a dimensionless quantity. The Euclidean distance mentioned in the homwork is just a way to specify "closest" when the answer has two numbers that we want to be close to simultaneously, so we just apply the Euclidean distance (square root of the sum of squared differences) as a way to quantify closeness into a single number.

[rsridharan]

Thank you very much Professer.