Quiz: Perceptron algorithm

Quiz: Perceptron algorithm

by Artur Szalata -
Number of replies: 1

We were told that the following answer is correct
"An increase of the length of the weight vector implies that the hyperplane does not change in N+1 dimensions"

While it seems to me that during an update we might have both, change of direction and length and so the implication from the quiz wouldn't hold. We would have the change of length, but also of direction, which would result in a changed hyperplane in both, N and N+1 dimensions.

May I ask for clarification?

In reply to Artur Szalata

Re: Quiz: Perceptron algorithm

by Bernd Albert Illing -

Hi, my interpretation of the question is that an update is considered which only changes the length of the weight vector but not its direction. In this case the hyperplane changes in N dimensions (distance of the hyperplane to the origin is changed) but not in N+1 dimensions (since it always goes through the origin in N+1 dimensions). I hope this helps.