[i]u[/i] is an eigenvector of matrix [i]A[/i] if its image through [i]A[/i] (i.e. [i]A*u[/i]), is collinear with [i]u[/i]. The corresponding eigenvalue -lambda- is the ratio of the (components of the) vectors [i]Au[/i] and [i]u[/i]. Because the vectors are collinear, the absolute value of lambda measures the ratio between the lengths of the vectors [i]Au[/i] and [i]u[/i].
Move the vector [i]u[/i] (drag the point [math]P[/math]) till the vectors [i]Au[/i] and [i]u[/i] becomes collinear. Then [i]u[/i] is an eigenvector of matrix [i]A[/i], lambda the corresponding eigenvalue.