Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...
Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...
Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
tel-00703797, version 2 - 7 Jun 2012<br />
3.2. Methods to solve <strong>non</strong>linear equations<br />
Therefore, the semismoothness of function H at x requires the Hadamard directional<br />
<strong>de</strong>rivative at x to exists along any direction converging to h and not only along h.<br />
Examples of semismooth functions are smooth, convex and piecewise linear functions. An<br />
important property is that composite, scalar products, sums, minimum, maximum preserve<br />
semismoothness. In<strong>de</strong>ed for the minimum function, the breakpoint is at (0, 0). By standard<br />
calculations, the directional <strong>de</strong>rivate of φ∧ at this point is φ∧((0, 0); (a, b)) = min(a, b).<br />
Futhermore, for all <strong>non</strong>zero vector (a, b), φ∧ is differentiable and is uniquely given by<br />
We <strong>de</strong>duce<br />
∂Bφ∧(0, 0) =<br />
<br />
1 0<br />
,<br />
0 1<br />
<br />
1a≤b<br />
∇φ∧(a, b) = .<br />
1a>b<br />
and ∂φ∧(0, 0) =<br />
Furthermore, for all V ∈ ∂φ∧(c, d), such that (c, d) → (a, b), we have<br />
V<br />
<br />
λ<br />
, λ ∈ [0, 1] . (3.17)<br />
1 − λ<br />
<br />
a<br />
− φ∧((0, 0); (a, b)) = a 1c≤d + b 1c>d − min(a, b) = o((a, b)) . (3.18)<br />
b<br />
By Appendix 3.5.1, we conclu<strong>de</strong> that φ∧ is semismooth at (0,0).<br />
Finally, we introduce the strong semismoothness, also called 1-or<strong>de</strong>r semismoothness, that<br />
will be used in most convergence theorems.<br />
Definition (strongly semismooth). A locally Lipschitzian function G is strongly semismooth<br />
at x if for all d → 0, ∀V ∈ ∂G(x + d), we have<br />
V d − G ′ (x, d) = O ||d|| 2 .<br />
Based on Equation (3.18), we conclu<strong>de</strong> that the minimum function φ∧ is not strongly<br />
semismooth at (0, 0). But, the Fischer-Burmeister function is strongly semismooth at (0, 0).<br />
In fact, by standard calculations, for all <strong>non</strong>zero vector (a, b), we have<br />
We <strong>de</strong>duce<br />
∇φF B(a, b) =<br />
<br />
√ a − 1<br />
a2 +b2 √ b − 1<br />
a2 +b2 ∂BφF B(0, 0) = {∇φF B(a, b), (a, b) = (0, 0)} and ∂φF B(0, 0) = ¯ B((−1, −1), 1), (3.19)<br />
where ¯ B <strong>de</strong>notes the closed ball.<br />
Furthermore, for all V ∈ ∂φF B(c, d), such that (c, d) → (a, b), we have<br />
Hence, we have<br />
V<br />
<br />
a<br />
= φF<br />
b<br />
B(a, b) and φF B((0, 0); (a, b)) = φF B(a, b).<br />
V<br />
<br />
<br />
a<br />
− φF<br />
b<br />
B((0, 0); (a, b)) = 0.<br />
.<br />
153