Skip to main content
. Author manuscript; available in PMC: 2016 Jul 1.
Published in final edited form as: Math Biosci. 2016 May 3;277:126–135. doi: 10.1016/j.mbs.2016.04.011

Table 5.

Column 1: numbering of eigenvectors. Column 2: Eigenvalues (smallest at the first row and then increasing values). Columns 3, 4 and 5: The value of the cost function J(r* + δri) where the parameter vector r* is given in Table 4 and ri denotes the i–th eigenvector. Recall that J(r*) = 0 so the table illustrates how much the cost function is changed by changing the the parameter vector along and eigenvector of the Hessian matrix.

i λi δ = 10−5 δ = 10−3 δ = 10−1
1 7.579e-15 3.157e-25 5.636e-18 5.607e-10
2 1.383e-14 9.316e-25 5.074e-19 5.047e-11
3 7.361e-14 3.084e-24 2.593e-16 2.551e-08
4 1.183e-13 2.095e-25 1.081e-17 9.346e-10
5 1.209e-13 3.153e-24 1.567e-17 1.586e-09
6 1.654e-13 4.465e-24 7.279e-17 7.309e-09
7 1.038e-11 5.191e-22 6.645e-18 1.648e-10
8 1.986e-10 9.933e-21 1.017e-16 2.242e-10
9 7.319e-10 3.659e-20 3.910e-16 1.943e-09
10 7.090e-09 3.545e-19 3.563e-15 2.011e-09
11 1.482e-06 7.410e-17 7.409e-13 3.662e-08
12 3.952e-06 1.976e-16 1.974e-12 1.943e-08
13 3.712e-05 1.856e-15 1.856e-11 1.965e-07
14 1.235e-03 6.174e-14 6.170e-10 5.963e-06
15 2.294e-02 1.147e-12 1.138e-08 6.443e-05
16 3.543e-01 1.771e-11 1.769e-07 1.539e-03