So in the lecture, the axes of the graphs, I assumed that w1, w2 are w_old, w_new as [w1, w2...] of the weight vector are values independent of one another.
Now taking the equations,
W_new = W_old - learning_rate*derivative(J(W_old, b))
and J(W_minima, b_minima) <= J(W, b).
Now I couldn’t exactly understand how the shape of contour plot is elliptical, this got me thinking whether this contour graph can be a 2D interpretation of the actual plot where b is along the Z-axis. Ik this is all just algebra at the end of the day, but can some please clarify this?
The elliptical shape is due to the magnitudes of the features.
No, I don’t think so.