Q7 No feature transform but do we add an x_0

[patrickjtierney]

Title says it all. Do we still add the column of 1's to X so that that the weight vector is of length 3, or not?

[patrickjtierney]

I tried it both ways, and strangely enough, it turns out that the number of wrongly classified inputs in each case is exactly equal whether you use the X0=1 term or not, even though the weights change quite a lot. I wasn't expecting that.

[Andrs]

Q7 is about plain linear regression and you should have a w0 parameter that corresponds to the constant bias.

[yaser]

Originally Posted by Andrs

Q7 is about plain linear regression and you should have a w0 parameter that corresponds to the constant bias.

Correct.