Thursday, March 15, 2018

Lagrange interpolation polynomials

Count Joseph Louis Lagrange, French mathematician was born in 1736 and died in 1813.

We look, in this paragraph, an expression of the polynomial with degree at most \(\mathfrak{n}\) taking the same values as a given function in \(\mathfrak{n} + 1\) given two, two distincts points, then to study the committed error by trying to minimize it.


1) Definition of polynomials \({L_k}\)

Let \(\mathfrak{n}\) be a natural integer then \({x_0},{x_1}, \ldots ,{x_n}\) \(\left( {\mathfrak{n} + 1} \right)\) complex two, two distincts.
We look \(\left( {\mathfrak{n} + 1} \right)\) polynomials \({L_0}, \ldots ,{L_n}\) all with degree at most \(\mathfrak{n}\), verifying the equalities :
$\forall \left( i,~j \right)\in {{\left[\!\left[ 0,n \right]\!\right]}^{2}},~{{L}_{i}}\left( {{x}_{j}} \right)={{\delta }_{i,j}},$
(where \({\delta _{i,\mathfrak{j}}}\) is the kronecker symbol defined by \({\delta _{i,\mathfrak{j}}} = \left\{ \begin{array}{l}0 \ {\rm{ si }} \ i \ne j\\1 \ {\rm{ si }} \ i = j\end{array} \right.\) ).
Let \(i\) a natural integer element of given \(\left[\kern-0.15em\left[ {0,n}
\right]\kern-0.15em\right]\). The polynomial \({L_i}\) has a degree at most \(\mathfrak{n}\) and admits the \(\mathfrak{n}\) complex two, two distincts \({x_j}\), \(\mathfrak{j} \ne i\), as roots. Then, necessairely, it exist a constant \(C\) such as
\({L_i} = C\prod\limits_{i \ne j} {(X - {x_j})} \) .
The equality\({L_i}\left( {{x_i}} \right) = 1\) gives then \(C = \frac{1}{{\prod\limits_{j \ne i} {\left( {{x_i} - {x_j}} \right)} }}\) and then necessairely \({L_i} = \prod\limits_{j \ne i} {\left( {\frac{{X - {x_\mathfrak{j}}}}{{{x_i} - {x_\mathfrak{j}}}}} \right)} \) .
Reciprocally, if for all \(i \in \left[\kern-0.15em\left[ {0,n}
\right]\kern-0.15em\right],\)\({L_i} = \prod\limits_{\mathfrak{j} \ne i} {\left( {\frac{{X - {x_\mathfrak{j}}}}{{{x_{\dot i}} - {x_\mathfrak{j}}}}} \right)} \) then \({L_i}\) est well defined because the \({x_{\dot j}}\) are two, two distincts, with degree \(\mathfrak{n}\) exactly and finally the polynomials \({L_i}\) verify clairly the duality equalities: $\forall \left( i,~\mathfrak{j} \right)\in {{\left[\!\left[ 0,n \right]\!\right]}^{2}},\text{ }{{L}_{i}}\left( {{x}_{j}} \right)={{\delta }_{i,j}}$.
Let \(n\) a natural integer then \({x_0},{x_1},...,{x_n}\), \(\left( {n + 1} \right)\) given complex two, two distincts.
It exists one and only one set, noted \({({L_i})_{0 \le i \le \mathfrak{n}}}\), with \(\left( {n + 1} \right)\) polynomials of degree at most \(n\) verifying :
$\forall \left( i,~j \right)\in {{\left[\!\left[ 0,n \right]\!\right]}^{2}},~{{L}_{i}}\left( {{x}_{i}} \right)={{\delta }_{i,j}}.$
Furthermore : $\forall \left( i,~j \right)\in {{\left[\!\left[ 0,n \right]\!\right]}^{2}},\text{ }{{L}_{i}}=\prod\limits_{j\ne i}{\left( \frac{X-{{x}_{j}}}{{{x}_{{\dot{i}}}}-{{x}_{j}}} \right)}$.

Sunday, March 4, 2018

The three-levels formula

1) Linear forms on \(\mathbb{R}\left[ X \right].\)

Let \(a\) be a real. We note \({\varphi _a}\) the application from \(\mathbb{R}\left[ X \right]\) in \(\mathbb{R}\) which to a polynomial \(P\) associates \(P\left( a \right)\).
The application \({\varphi _a}\) is called the evaluation in \(a.\)
For all real \(a,\) \({\varphi _a}\) is a linear form on \(\mathbb{R}\left[ X \right]\). In fact, if \(P\) and \(Q\) are two polynomials and \(\lambda \) and \(\mu \) two reals, we have
\({\varphi _a}\left( {\lambda P + \mu Q} \right) = \left( {\lambda P + \mu Q} \right)\left( a \right) = \lambda P\left( a \right) + \mu Q\left( a \right) = \lambda {\varphi _a}\left( P \right) + \mu {\varphi _a}\left( Q \right)\) .

Let \(a\) be a real and \({\varphi _a}:\) \(\begin{array}{c}\mathbb{R}\left[ X \right] \to \mathbb{R}\\{\rm{ }}P \mapsto P\left( a \right)\end{array}\). \({\varphi _a}\) is a linar form on \(\mathbb{R}\left[ X \right].\)

On the other hand, Let \(a\) and \(b\) be two reals. We note furthermore \(\psi \) the application from \(\mathbb{R}\left[ X \right]\)in \(\mathbb{R}\)which to a polynomial \(P\) associates \(\int_a^b {P\left( t \right)dt} \). We know \(\psi \) is a linear form on \(\mathbb{R}\left[ X \right].\)