Difference between revisions of "Weighted Least Squares (WLS)"

From Medusa: Coordinate Free Mehless Method implementation
Jump to: navigation, search
(Shape functions :: WLS at fixed point with fixed support and unknown function values)
m (Typos)
 
(17 intermediate revisions by 3 users not shown)
Line 1: Line 1:
One of the most important building blocks of the meshless methods is the Moving Least Squares (MLS) approximation , which is implemented in the [http://www-e6.ijs.si/ParallelAndDistributedSystems/MeshlessMachine/technical_docs/html/classEngineMLS.html EngineMLS class]. Check [https://gitlab.com/e62Lab/e62numcodes/blob/master/test/mls_test.cpp EngineMLS unit tests] for examples.
+
One of the most important building blocks of the meshless methods is the Weighted/Moving Least Squares (WLS/MLS) approximation , which is implemented in the <code>WLS</code> class.
 +
It is also often called GFDM (generalized finite differences method).
  
= Notation Cheat sheet =
+
= Notation cheat sheet =
 
\begin{align*}
 
\begin{align*}
 
   m \in \N                  & \dots \text{number of basis functions} \\
 
   m \in \N                  & \dots \text{number of basis functions} \\
Line 10: Line 11:
 
   \vec p \in \R^k          & \dots \text{center point of approximation} \\
 
   \vec p \in \R^k          & \dots \text{center point of approximation} \\
 
   b_i\colon \R^k \to \R    & \dots \text{basis functions } \quad i=1,\dots,m \\
 
   b_i\colon \R^k \to \R    & \dots \text{basis functions } \quad i=1,\dots,m \\
   B_{j, i} \in \R          & \dots \text{value of basis functions in support points } b_i(s_j-p) \quad j=1,\dots,n, \quad i=1,\dots,m\\
+
   B_{j, i} \in \R          & \dots \text{value of basis functions in support points } b_i(\vec{s}_j) \quad j=1,\dots,n, \quad i=1,\dots,m\\
 
   \omega \colon \R^k \to \R & \dots \text{weight function} \\
 
   \omega \colon \R^k \to \R & \dots \text{weight function} \\
 
   w_j \in \R                & \dots \text{weights } \omega(\vec{s}_j-\vec{p})  \quad j=1,\dots,n \\
 
   w_j \in \R                & \dots \text{weights } \omega(\vec{s}_j-\vec{p})  \quad j=1,\dots,n \\
Line 18: Line 19:
 
\end{align*}
 
\end{align*}
  
We will also use \(\b{s}, \b{u}, \b{b}, \b{\alpha}, \b{\chi} \) to annotate a column of corresponding values,
+
We will also use \(\b{s}, \b{u}, \b{b}, \b{\alpha}, \b{\chi} \) to denote a column of corresponding values,
$W$ as a $n\times n$ diagonal matrix filled with $w_j$ on the diagonal and $B$ as a $n\times m$ matrix filled with $B_{j, i}$.
+
$W$ as a $n\times n$ diagonal matrix filled with $w_j$ on the diagonal and $B$ as a $n\times m$ matrix filled with $B_{j, i} = b_i(\vec s_j)$.
  
 
= Definition of local approximation =
 
= Definition of local approximation =
Line 63: Line 64:
  
  
The problem of finding the coefficients $\b{\alpha}$ that minimize the error $\b{e}$ can be solved with at least three approaches:
+
The problem of finding the coefficients $\b{\alpha}$ that minimize the error $\b{e}$ numerically is described in [[Solving overdetermined systems]].
* Normal equations (fastest, less accurate) - using Cholesky decomposition of $B$ (requires full rank and $m \leq n$)
 
* QR decomposition of $B$ (requires full rank and $m \leq n$, more precise)
 
* SVD decomposition of $B$ (more expensive, even more reliable, no rank demand)
 
  
In our Meshless Machine MLS engine we use SVD with regularization described below.
+
Here, we simply write the solution $\b \alpha = (WB)^+ Wu = (B^\T W^2 B)^{-1} B^\T W^2 u$.
  
= Computing approximation coefficients =
+
In our WLS engine we use SVD with regularization by default.
 
 
== [http://mathworld.wolfram.com/NormalEquation.html Normal equations] ==
 
We seek the minimum of
 
\[ \|\b{e}\|_2^2 = (B\b{\alpha} - \b{u})^\T(B\b{\alpha} - \b{u}) \]
 
By seeking the zero gradient in terms of coefficients $\alpha_i$
 
\[\frac{\partial}{\partial \alpha_i} (B\b{\alpha} - \b{u})^\T(B\b{\alpha} - \b{u})  = 0\]
 
resulting in
 
\[ B^\T B\b{\alpha} = B^\T \b{u}. \]
 
The coefficient matrix $B^\T B$ is symmetric and positive definite. However, solving above problem directly is
 
poorly behaved with respect to round-off errors since the condition number $\kappa(B^\T B)$ is the square
 
of $\kappa(B)$.
 
 
 
In case of WLS the equations become
 
\[ (WB)^\T WB \b{\alpha} = WB^\T \b{u}. \]
 
 
 
Complexity of Cholesky decomposition is $\frac{n^3}{3}$ and complexity of matrix multiplication $nm^2$. To preform the Cholesky decomposition the rank of $WB$ must be full.
 
 
 
'''Pros:'''
 
* simple to implement
 
* low computational complexity
 
 
 
'''Cons:'''
 
* numerically unstable
 
* full rank requirement
 
 
 
== [https://en.wikipedia.org/wiki/QR_decomposition $QR$ Decomposition] ==
 
\[{\bf{B}} = {\bf{QR}} = \left[ {{{\bf{Q}}_1},{{\bf{Q}}_2}} \right]\left[ {\begin{array}{*{20}{c}}
 
{{{\bf{R}}_1}}\\
 
0
 
\end{array}} \right]\]
 
\[{\bf{B}} = {{\bf{Q}}_1}{{\bf{R}}_1}\]
 
$\bf{Q}$ is unitary matrix ($\bf{Q}^{-1}=\bf{Q}^T$). Useful property of a unitary matrices is that multiplying with them does not alter the (Euclidean) norm of a vector, i.e.,
 
\[\left\| {{\bf{Qx}}} \right\|{\bf{ = }}\left\| {\bf{x}} \right\|\]
 
And $\bf{R}$ is upper diagonal matrix
 
\[{\bf{R = (}}{{\bf{R}}_{\bf{1}}}{\bf{,}}0{\bf{)}}\]
 
therefore we can say
 
\[\begin{array}{l}
 
\left\| {{\bf{B\alpha }} - {\bf{u}}} \right\| = \left\| {{{\bf{Q}}^{\rm{T}}}\left( {{\bf{B\alpha }} - {\bf{u}}} \right)} \right\| = \left\| {{{\bf{Q}}^{\rm{T}}}{\bf{B\alpha }} - {{\bf{Q}}^{\rm{T}}}{\bf{u}}} \right\|\\
 
= \left\| {{{\bf{Q}}^{\rm{T}}}\left( {{\bf{QR}}} \right){\bf{\alpha }} - {{\bf{Q}}^{\rm{T}}}{\bf{u}}} \right\| = \left\| {\left( {{{\bf{R}}_1},0} \right){\bf{\alpha }} - {{\left( {{{\bf{Q}}_1},{{\bf{Q}}_{\bf{2}}}} \right)}^{\rm{T}}}{\bf{u}}} \right\|\\
 
= \left\| {{{\bf{R}}_{\bf{1}}}{\bf{\alpha }} - {{\bf{Q}}_{\bf{1}}}{\bf{u}}} \right\| + \left\| {{\bf{Q}}_2^{\rm{T}}{\bf{u}}} \right\|
 
\end{array}\]
 
Of the two terms on the right we have no control over the second, and we can render the first one
 
zero by solving
 
\[{{\bf{R}}_{\bf{1}}}{\bf{\alpha }} = {\bf{Q}}_{_{\bf{1}}}^{\rm{T}}{\bf{u}}\]
 
Which results in a minimum. We could also compute it with pseudo inverse
 
\[\mathbf{\alpha }={{\mathbf{B}}^{-1}}\mathbf{u}\]
 
Where pseudo inverse is simply \[{{\mathbf{B}}^{-1}}=\mathbf{R}_{\text{1}}^{\text{-1}}{{\mathbf{Q}}^{\text{T}}}\] (once again, $R$ is upper diagonal matrix, and $Q$ is unitary matrix).
 
And for weighted case
 
\[\mathbf{\alpha }={{\left( {{\mathbf{W}}^{0.5}}\mathbf{B} \right)}^{-1}}\left( {{\mathbf{W}}^{0.5}}\mathbf{u} \right)\]
 
 
 
Complexity of $QR$ decomposition \[\frac{2}{3}m{{n}^{2}}+{{n}^{2}}+\frac{1}{3}n-2=O({{n}^{3}})\]
 
 
 
<strong>Pros:</strong> better stability in comparison with normal equations <strong> cons: </strong>higher complexity
 
 
 
== [https://en.wikipedia.org/wiki/Singular_value_decomposition SVD decomposition] ==
 
In linear algebra, the [https://en.wikipedia.org/wiki/Singular_value_decomposition singular value decomposition (SVD)]
 
is a factorization of a real or complex matrix. It has many useful
 
applications in signal processing and statistics.
 
 
 
Formally, the singular value decomposition of an $m \times n$ real or complex
 
matrix $\bf{B}$ is a factorization of the form $\bf{B}= \bf{U\Sigma V^\T}$, where
 
$\bf{U}$ is an $m \times m$ real or complex unitary matrix, $\bf{\Sigma}$ is an $m \times n$
 
rectangular diagonal matrix with non-negative real numbers on the diagonal, and
 
$\bf{V}^\T$  is an $n \times n$ real or complex unitary matrix. The diagonal entries
 
$\Sigma_{ii}$ are known as the singular values of $\bf{B}$. The $m$ columns of
 
$\bf{U}$ and the $n$ columns of $\bf{V}$ are called the left-singular vectors and
 
right-singular vectors of $\bf{B}$, respectively.
 
 
 
The singular value decomposition and the eigen decomposition are closely
 
related. Namely:
 
 
 
* The left-singular vectors of $\bf{B}$ are eigen vectors of $\bf{BB}^\T$.
 
* The right-singular vectors of $\bf{B}$ are eigen vectors of $\bf{B}^\T{B}$.
 
* The non-zero singular values of $\bf{B}$ (found on the diagonal entries of $\bf{\Sigma}$) are the square roots of the non-zero eigenvalues of both $\bf{B}^\T\bf{B}$ and $\bf{B}^\T \bf{B}$.
 
 
 
with SVD we can write $\bf{B}$ as \[\bf{B}=\bf{U\Sigma{{V}^{\T}}}\] where are $\bf{U}$ and $\bf{V}$ again unitary matrices and $\bf{\Sigma}$
 
stands for diagonal matrix of singular values.
 
 
 
Again we can solve either the system or compute pseudo inverse as
 
 
 
\[ \bf{B}^{-1} = \left( \bf{U\Sigma V}^\T\right)^{-1} = \bf{V}\bf{\Sigma^{-1}U}^\T \]
 
where $\bf{\Sigma}^{-1}$ is trivial, just replace every non-zero diagonal entry by
 
its reciprocal and transpose the resulting matrix. The stability gain is
 
exactly here, one can now set threshold below which the singular value is
 
considered as $0$, basically truncate all singular values below some value and
 
thus stabilize the inverse.
 
 
 
SVD decomposition complexity \[ 2mn^2+2n^3 = O(n^3) \]
 
 
 
<strong>Pros:</strong> stable <strong> cons: </strong>high complexity
 
 
 
Method used in MLSM is SVD with regularization.
 
  
 
= Weighted Least Squares =
 
= Weighted Least Squares =
Line 176: Line 82:
 
constant and applied the linearity.
 
constant and applied the linearity.
 
\[
 
\[
  \widehat{\mathcal L u}(\vec x) = \sum_{i=1}^m \alpha_i (\mathcal L b_i)(\vec x).
+
  \widehat{\mathcal L u}(\vec x) = \sum_{i=1}^m \alpha_i (\mathcal L b_i)(\vec x) = (\mathcal L \b b(\vec x))^\T \b \alpha.
 
\]
 
\]
  
= MLS =
+
= Moving Least Squares =
  
 
<figure id="fig:comparisonMLSandWLS">
 
<figure id="fig:comparisonMLSandWLS">
Line 197: Line 103:
 
the weighted least squares problem whenever we move the point of approximation.
 
the weighted least squares problem whenever we move the point of approximation.
  
Note that if out weight is constant or if $n = m$, when approximation reduces to interpolation, the weights do not play
+
Note that if our weight is constant or if $n = m$, when approximation reduces to interpolation, the weights do not play
 
any role and this method is redundant. In fact, its benefits arise when supports are rather large.
 
any role and this method is redundant. In fact, its benefits arise when supports are rather large.
  
 
See <xr id="fig:comparisonMLSandWLS"/> for comparison between MLS and WLS approximations. MLS approximation remains close to
 
See <xr id="fig:comparisonMLSandWLS"/> for comparison between MLS and WLS approximations. MLS approximation remains close to
actual function while still inside the support domain, while WLS approximation becomes bad when
+
the actual function while still inside the support domain, while WLS approximation becomes bad when
 
we come out of the reach of the weight function.
 
we come out of the reach of the weight function.
  
 
{{reflist}}
 
{{reflist}}

Latest revision as of 13:57, 16 June 2022

One of the most important building blocks of the meshless methods is the Weighted/Moving Least Squares (WLS/MLS) approximation , which is implemented in the WLS class. It is also often called GFDM (generalized finite differences method).

Notation cheat sheet

\begin{align*} m \in \N & \dots \text{number of basis functions} \\ n \geq m \in \N & \dots \text{number of points in support domain} \\ k \in \mathbb{N} & \dots \text{dimensionality of vector space} \\ \vec s_j \in \R^k & \dots \text{point in support domain } \quad j=1,\dots,n \\ u_j \in \R & \dots \text{value of function to approximate in }\vec{s}_j \quad j=1,\dots,n \\ \vec p \in \R^k & \dots \text{center point of approximation} \\ b_i\colon \R^k \to \R & \dots \text{basis functions } \quad i=1,\dots,m \\ B_{j, i} \in \R & \dots \text{value of basis functions in support points } b_i(\vec{s}_j) \quad j=1,\dots,n, \quad i=1,\dots,m\\ \omega \colon \R^k \to \R & \dots \text{weight function} \\ w_j \in \R & \dots \text{weights } \omega(\vec{s}_j-\vec{p}) \quad j=1,\dots,n \\ \alpha_i \in \R & \dots \text{expansion coefficients around point } \vec{p} \quad i=1,\dots,m \\ \hat u\colon \R^k \to \R & \dots \text{approximation function (best fit)} \\ \chi_j \in \R & \dots \text{shape coefficient for point }\vec{p} \quad j=1,\dots,n \\ \end{align*}

We will also use \(\b{s}, \b{u}, \b{b}, \b{\alpha}, \b{\chi} \) to denote a column of corresponding values, $W$ as a $n\times n$ diagonal matrix filled with $w_j$ on the diagonal and $B$ as a $n\times m$ matrix filled with $B_{j, i} = b_i(\vec s_j)$.

Definition of local approximation

1D MLS example
Figure 1: Example of 1D WLS approximation

Our wish is to approximate an unknown function $u\colon \R^k \to \R$ while knowing $n$ values $u(\vec{s}_j) := u_j$. The vector of known values will be denoted by $\b{u}$ and the vector of coordinates where those values were achieved by $\b{s}$. Note that $\b{s}$ is not a vector in the usual sense since its components $\vec{s}_j$ are elements of $\R^k$, but we will call it vector anyway. The values of $\b{s}$ are called nodes or support nodes or support. The known values $\b{u}$ are also called support values.

In general, an approximation function around point $\vec{p}\in\R^k$ can be written as \[\hat{u} (\vec{x}) = \sum_{i=1}^m \alpha_i b_i(\vec{x}) = \b{b}(\vec{x})^\T \b{\alpha} \] where $\b{b} = (b_i)_{i=1}^m$ is a set of basis functions, $b_i\colon \R^k \to\R$, and $\b{\alpha} = (\alpha_i)_{i=1}^m$ are the unknown coefficients.

In MLS the goal is to minimize the error of approximation in given values, $\b{e} = \hat u(\b{s}) - \b{u}$ between the approximation function and target function in the known points $\b{x}$. The error can also be written as $B\b{\alpha} - \b{u}$, where $B$ is rectangular matrix of dimensions $n \times m$ with rows containing basis function evaluated in points $\vec{s}_j$. \[ B = \begin{bmatrix} b_1(\vec{s}_1) & \ldots & b_m(\vec{s}_1) \\ \vdots & \ddots & \vdots \\ b_1(\vec{s}_n) & \ldots & b_m(\vec{s}_n) \end{bmatrix} = [b_i(\vec{s}_j)]_{j=1,i=1}^{n,m} = [\b{b}(\vec{s}_j)^\T]_{j=1}^n. \]

We can choose to minimize any norm of the error vector $e$ and usually choose to minimize the $2$-norm or square norm \[ \|\b{e}\| = \|\b{e}\|_2 = \sqrt{\sum_{j=1}^n e_j^2}. \] Commonly, we also choose to minimize a weighted norm [1] instead \[ \|\b{e}\|_{2,w} = \|\b{e}\|_w = \sqrt{\sum_{j=1}^n (w_j e_j)^2}. \] The weights $w_i$ are assumed to be non negative and are assembled in a vector $\b{w}$ or a matrix $W = \operatorname{diag}(\b{w})$ and usually obtained from a weight function. A weight function is a function $\omega\colon \R^k \to[0,\infty)$. We calculate $w_j$ as $w_i := \omega(\vec{p}-\vec{s}_j)$, so good choices for $\omega$ are function which have higher values close to $0$ (making closer nodes more important), like the normal distribution. If we choose $\omega \equiv 1$, we get the unweighted version.

A choice of minimizing the square norm gave this method its name - Least Squares approximation. If we use the weighted version, we get the Weighted Least Squares or WLS. In the most general case we wish to minimize \[ \|\b{e}\|_{2,w}^2 = \b{e}^\T W^2 \b{e} = (B\b{\alpha} - \b{u})^\T W^2(B\b{\alpha} - \b{u}) = \sum_j^n w_j^2 (\hat{u}(\vec{s}_j) - u_j)^2 \]


The problem of finding the coefficients $\b{\alpha}$ that minimize the error $\b{e}$ numerically is described in Solving overdetermined systems.

Here, we simply write the solution $\b \alpha = (WB)^+ Wu = (B^\T W^2 B)^{-1} B^\T W^2 u$.

In our WLS engine we use SVD with regularization by default.

Weighted Least Squares

Weighted least squares approximation is the simplest version of the procedure described above. Given support $\b{s}$, values $\b{u}$ and an anchor point $\vec{p}$, we calculate the coefficients $\b{\alpha}$ using one of the above methods. Then, to approximate a function in the neighbourhood of $\vec p$ we use the formula \[ \hat{u}(\vec x) = \b{b}(\vec x)^\T \b{\alpha} = \sum_{i=1}^m \alpha_i b_i(\vec x). \]

To approximate the derivative $\frac{\partial u}{\partial x_i}$, or any linear partial differential operator $\mathcal L$ on $u$, we simply take the same linear combination of transformed basis functions $\mathcal L b_i$. We have considered coefficients $\alpha_i$ to be constant and applied the linearity. \[ \widehat{\mathcal L u}(\vec x) = \sum_{i=1}^m \alpha_i (\mathcal L b_i)(\vec x) = (\mathcal L \b b(\vec x))^\T \b \alpha. \]

Moving Least Squares

Figure 2: Comparison of WLS and MLS approximation

When using WLS the approximation gets worse as we move away from the central point $\vec{p}$. This is partially due to not being in the center of the support any more and partially due to weight being distributed in such a way to assign more importance to nodes closer to $\vec{p}$.

We can battle this problem in two ways: when we wish to approximate in a new point that is sufficiently far away from $\vec{p}$ we can compute new support, recompute the new coefficients $\b{\alpha}$ and approximate again. This is very costly and we would like to avoid that. A partial fix is to keep support the same, but only recompute the weight vector $\b{w}$, that will now assign weight values to nodes close to the new point. We still need to recompute the coefficients $\b{\alpha}$, however we avoid the cost of setting up new support and function values and recomputing $B$. This approach is called Moving Least Squares due to recomputing the weighted least squares problem whenever we move the point of approximation.

Note that if our weight is constant or if $n = m$, when approximation reduces to interpolation, the weights do not play any role and this method is redundant. In fact, its benefits arise when supports are rather large.

See Figure 2 for comparison between MLS and WLS approximations. MLS approximation remains close to the actual function while still inside the support domain, while WLS approximation becomes bad when we come out of the reach of the weight function.

End notes

  1. Note that our definition is a bit unusual, usually weights are not squared with the values. However, we do this to avoid computing square roots when doing MLS. If you are used to the usual definition, consider the weight to be $\omega^2$.