Question on Gradient

[melipone]

For gradient descent in Q5-6, is the gradient a vector consisting of the partial derivatives wrt u and the partial derivative wrt v or is it the derivative wrt to both u and v?

[kartikeya_t@yahoo.com]

It is the former, i.e., a vector whose first component is the partial derivative of E wrt u, and the second component is the partial derivative of E wrt v.

[melipone]

Got it!
I have another question: Are you supposed to update the value of the point after updating the weights? I don't see that in the algorithm on p. 95.

[yaser]

Originally Posted by melipone

Got it!
I have another question: Are you supposed to update the value of the point after updating the weights? I don't see that in the algorithm on p. 95.

In this specific problem, the "weights" are , namely the parameters you are optimizing with respect to.

[melipone]

ah, the point (u,v) = (1,1) is the initial weight?

[butterscotch]

you are right!