mathematics

Article
A Matlab Toolbox for Extended Dynamic Mode Decomposition
Based on Orthogonal Polynomials and p-q Quasi-Norm
Order Reduction
Camilo Garcia-Tenorio and Alain Vande Wouwer *

Systems, Estimation, Control and Optimization (SECO), Université de Mons, 7000 Mons, Belgium
* Correspondence: [email protected]

Abstract: Extended Dynamic Mode Decomposition (EDMD) allows an approximation of the Koop-
man operator to be derived in the form of a truncated (finite dimensional) linear operator in a lifted
space of (nonlinear) observable functions. EDMD can operate in a purely data-driven way using
either data generated by a numerical simulator of arbitrary complexity or actual experimental data.
An important question at this stage is the selection of basis functions to construct the observable
functions, which in turn is determinant of the sparsity and efficiency of the approximation. In this
study, attention is focused on orthogonal polynomial expansions and an order-reduction procedure
called p-q quasi-norm reduction. The objective of this article is to present a Matlab library to automate
the computation of the EDMD based on the above-mentioned tools and to illustrate the performance
of this library with a few representative examples.

Citation: Garcia-Tenorio, C.; Vande Keywords: extended dynamic mode decomposition; Koopman operator; orthogonal polynomials;
Wouwer, A. A Matlab Toolbox for mathematical modeling; dynamic systems
Extended Dynamic Mode
Decomposition Based on Orthogonal MSC: 37-04
Polynomials and p-q Quasi-Norm
Order Reduction. Mathematics 2022,
10, 3859. https://doi.org/10.3390/
math10203859 1. Introduction
Academic Editors: Francesco Aldo In contrast to traditional modeling approaches, in which it is necessary to formulate a
Costabile, Maria I. Gualtieri, Anna general nonlinear model that depends on a set of parameters to replicate the dynamics of
Napoli, Stefano De Marchi and a system under consideration, the EDMD [1] builds upon numerical data (simulation or
Xiangmin Jiao actual experiments) to provide a finite-dimensional (truncated) linear representation of the
system dynamics in a lifted space of nonlinear observable functions, making it akin to a
Received: 31 August 2022
black box modeling paradigm, e.g., transfer functions or autoregressive models [2–4]. While
Accepted: 14 October 2022
the approximation provided by EDMD therefore remains nonlinear in the original state
Published: 18 October 2022
variables, it is linear in a transformed space, which is called the observable space, function
Publisher’s Note: MDPI stays neutral space, or vector-valued function of observables (VVFO), among others (for the remainder
with regard to jurisdictional claims in of this article, we use observables when referring to this space, while the example codes use
published maps and institutional affil- the VVFO terminology). In other words, The EDMD formulation does not represent the
iations. system by a linearized representation of the form x (k + 1) = Ax (k) (where A is a Jacobian);
rather, it describes the evolution of observables f ( x (k)) through a linear operator U, i.e.,

f ( x (k + 1)) = U f ( x (k )). (1)

Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland. EDMD is closely related to other decompositions such as Karhunen–Loeve decomposi-
This article is an open access article tion (KLD) [5], singular value decomposition (SVD) [6], proper orthogonal decomposition
distributed under the terms and
(POD) [7], and its direct precursor, dynamic mode decomposition (DMD) [8]. These de-
conditions of the Creative Commons
compositions all produce linear approximations of the behavior of the system near a fixed
Attribution (CC BY) license (https://
point, offering the possibility of using linear system analysis tools. EDMD extends this
creativecommons.org/licenses/by/
possibility to a region of the state space which is larger than the neighborhood of the fixed
4.0/).

Mathematics 2022, 10, 3859. https://doi.org/10.3390/math10203859 https://www.mdpi.com/journal/mathematics

Mathematics 2022, 10, 3859 2 of 18

point. Indeed, EDMD describes the nonlinear dynamics while being linear in the function
space, and therefore provides more than local information while preserving the linear
characteristics of the above-mentioned decompositions (KLD, SVD, POD, DMD). As such,
it is sometimes called a linearization in the large.
Koopman mode decomposition (KMD) [9–11] emerges from the linearity of decompo-
sitions that use their spectrum (eigenvalues and eigenvectors) to obtain an approximation of
the Koopman operator [12]. From an approximation of the Koopman operator, it is possible
to analyze nonlinear systems in terms of their stability and regions of attraction [9,13–15].
Additionally, EDMD (or the approximate Koopman operator) can be used in the context
of optimal control and model predictive control [16–19]. These developments show the
importance of having accurate EDMD approximations for analysis and control.
There are several variants of the EDMD algorithm, which use norm-based expansions,
radial-basis functions, kernel-based representations [20], orthogonal polynomials, and their
variations [21,22]. These representations provide tools for analyzing nonlinear systems
via spectral decomposition, and represent the fundamentals for developing synthesis
algorithms such as EDMD for control [23].
In this paper, attention is focused on the use of orthogonal polynomials for the ex-
pression of the observable functions and an order reduction method based on p-q quasi
norm [24,25]. Several application examples are described together with Matlab codes which
constitute a practical library for users interested in applying EDMD to engineering and
scientific problems.
The library provides several Matlab functions to compute a pq-EDMD approximation
of a dynamical system based on one of three possible algorithms. The first algorithm is the
original least squares solution, which is suitable for data with a high signal-to-noise ratio.
For data with higher levels of noise, a maximum likelihood approximation is proposed,
which is valid for unimodal Gaussian distributions (i.e., Gaussian noise for a system with a
unique stable equilibrium point). Finally, a solution based on regularized least squares is
provided, which promotes sparsity in the regression matrix.

2. Extended Dynamic Mode Decomposition

This section starts with an introduction to the traditional EDMD formulation to identify
nonlinear models of dynamical systems. The procedure is exemplified by the Duffing
equation, a benchmark problem in the literature for testing the reliability of the algorithm.
The core idea of the EDMD algorithm is to transform a nonlinear system into an
augmented linear system. The first proponent of this idea was Takata [26], who describes
the method as a formal linearization. Much later, the method emerged in its current form
after the development of the dynamic mode decomposition algorithm [8] and its several
extensions. In the following, the original EDMD algorithm [1] is first presented, followed
by pq-EDMD, which makes use of orthogonal polynomials and order reduction based on
p-q quasi-norm. This later version is particularly interesting as it yields increased numerical
accuracy and systematic application.

2.1. The Basic EDMD Formulation

Consider as an example the unforced Duffing oscillator, which is a nonlinear spring
that has different behaviors depending on the parameterization. The set of differential
equations that govern this system is

ẋ1 = x2 (2)
ẋ2 = −δx2 − x1 (α + βx12 ). (3)

where state x1 is the displacement, state x2 is the velocity, α is the stiffness of the spring,
which is related to the linear force of the spring according to Hooke’s law, δ is the amount
of damping in the system, and β is the proportion of nonlinearity present in the stiffness of
the spring.
Mathematics 2022, 10, 3859 3 of 18

Figure 1 shows the phase plane of the system for three different sets of parameters and
six random initial conditions (i.e., six initial conditions are generated randomly within the
range ics ∈ [−2, 2]2 starting from a known seed rng(1)). This choice of initial conditions
produces an appropriate set of trajectories for calculating the approximation and testing
its accuracy. The system of Equations (2) and (3) is integrated with a Matlab ODE solver,
e.g., ode23s, and the results are collected at a constant sample period ∆t = 0.1 s for a total of
20 s. The result of this numerical integration is a set of six trajectories of two state variables
with 201 points per variable. Each of these trajectories is an element of a structure array in
Matlab with the fields “Time” and “SV”. The choice of a structure array instead of a tensor
comes from the possibility of having trajectories of different lengths, e.g., experimental data
of different lengths, a feature that becomes important in systems where having redundant
data near the asymptotically stable attractors has a negative impact on the approximation.

Figure 1. Orbits of the Duffing equation for different parameterizations. (Left): undamped, (Center):
hard spring, (Right): soft spring, (Top row): phase plane, (Bottom): states versus time of two
trajectories.

As the EDMD is a data-driven algorithm, certain trajectories serve as a training set

while others serve as a testing set. The amount of data necessary to obtain an accurate
approximation depends on the system under consideration as well as its information
content (large data sets can bear little information content if experiments are not properly
designed). The EDMD algorithm captures the dynamic of the system on the portion of the
state space covered by the trajectories in the training set. Therefore, designing experiments
that maximize the coverage of the state space can reduce the amount of data while having
a positive effect on accuracy.
Each trajectory of the training set is in discrete-time, i.e., x (k + 1) = T ( x (k )), where
x ∈ Rn are the states of the system, k ∈ Z0+ is the non-negative discrete time, and
T : Rn → Rn is an unknown nonlinear mapping that provides the evolution of the discrete-
time trajectories. To construct the database, the training trajectories are organized in
so-called snapshot pairs {( xi , yi )}iN=1 , where yi = T ( xi ). The snapshots function presented
in Listing 1 handles the available trajectories, dividing them into training and testing sets
of the appropriate type; the training set consists of matrices containing the x and y data,
while the testing set is a cell array containing one orbit per index of the cell. The choice of
cell arrays instead of a tensor is to offer the possibility of testing trajectories of different
lengths. The tr_ts argument is a Matlab structure containing the indexes of the original
set of orbits, which serve as the training and testing sets. The fields of this structure must
Mathematics 2022, 10, 3859 4 of 18

be tr_index and ts_index, respectively. In addition, there is a normalization flag to use

when necessary, e.g., when the order of magnitude of different states is dissimilar.

Listing 1. Function snapshots used to create the data pairs for training and testing.

1
2 function [ xtr , ytr , xts , yts , center , scale ] = snapshots (
system , tr_ts , normalization )
3 % Number of trajectories from the system to populate the
4 % snapshots
5 training_number = numel ( tr_ts . tr_index ) ;
6 % First , store the snapshots in a cell
7 [ xtr_cell , ytr_cell ] = deal ( cell ( training_number ,1) ) ;
8 for trj = 1 : training_number
9 % Extract the appropriate data points for x and y
10 xtr_cell { trj } = system ( tr_ts . tr_index ( trj ) ) . SV (1: end
-2 ,:) ;
11 ytr_cell { trj } = system ( tr_ts . tr_index ( trj ) ) . SV (2: end
-1 ,:) ;
12 end
13 % Turn cells into a matrix
14 xtr = cell2mat ( xtr_cell ) ;
15 ytr = cell2mat ( ytr_cell ) ;
16 % Normalize if necessary
17 if normalization
18 [ xtr , center , scale ] = normalize ( xtr , ` ` zscore ' ') ;
19 ytr = normalize ( ytr , ' center ' , center , ' scale ' , scale ) ;
20 else
21 center = zeros (1 , size ( xtr ,2) ) ;
22 scale = ones (1 , size ( xtr ,2) ) ;
23 end
24 % For the test trajectories , we need each of them in a
25 % differerent cell because we want to compare the whole
26 % trajectory based on the initial condition
27 testing_number = numel ( tr_ts . ts_index ) ;
28 [ xts , yts ] = deal ( cell ( testing_number ,1) ) ;
29 for trj = 1 : testing_number
30 xts { trj } = normalize ( system ( tr_ts . ts_index ( trj ) ) . SV (1:
end -2 ,:) , ...
31 ' center ' , center (1: size ( system ( tr_ts . ts_index ( trj ) ) .
SV ,2) ) , ...
32 ' scale ' , scale (1: size ( system ( tr_ts . ts_index ( trj ) ) . SV
,2) ) ) ;
33 yts { trj } = normalize ( system ( tr_ts . ts_index ( trj ) ) . SV (2:
end -1 ,:) , ...
34 ' center ' , center (1: size ( system ( tr_ts . ts_index ( trj ) ) .
SV ,2) ) , ...
35 ' scale ' , scale (1: size ( system ( tr_ts . ts_index ( trj ) ) . SV
,2) ) ) ;
36 end
37 end

Notice that the generation of the snapshots avoids the last element in each trajectory,
SV(1:end-2) for x and SV(2:end-1) for y. As stated before, avoiding redundant data at
the asymptotically stable attractors improves the performance of the algorithm. In Matlab,
stopping the simulation early, e.g., as convergence towards the attractor has been achieved,
causes the last output interval ∆t 6= 0.1. This small difference can increase the error in the
Mathematics 2022, 10, 3859 5 of 18

construction of the approximation. Conversely, if the numerical integration of the system

is not stopped near the steady state, it is not necessary to eliminate the last element in
the trajectories.
The next step in the development of the EDMD is the definition of the observable
space as a set of functions f i ( x ) : Rn → C for i = 1, . . . , d, which represent a transformation
from the state space into an arbitrary function space. This transformation of the state
is equivalent to a change of variables z = f ( x ), where z ∈ Cd . In the Matlab library,
the observables are described by orthogonal polynomials, where each element of the set
of observables is the tensor product of n univariate polynomials up to order p ∈ N+ .
For example, in the Duffing oscillator, a set of observables with p = 2 and a Hermite basis
of orthogonal polynomials is provided by

f ( x ) = 1 2x1 4x12 − 2


2x2 4x1 x2 2x2 (4x12 − 2)

>
4x22 − 2 2x1 (4x22 − 2) (4x12 − 2)(4x22 − 2) . (4)

Note that the first entry is the product of a zero-order polynomial in both of the state
variables; the orders of the polynomial basis in the two state variables can be summarized by

x1 : 0 1 2 0 1 2 0 1 2
(5)
x2 : 0 0 0 1 1 1 2 2 2

making the generation of observables a problem of accurately handling indexes. Notice

that the full basis of indexes (5) is equivalent to counting numbers on a p + 1 basis with n
significant figures. From such a set of indexes, a method to generate a set of observables
with a Hermite base is proposed in Listing 2.

Listing 2. Generation of a set of observables with a Hermite basis.

1 % Generate the matrix p of indexes

2 hpm = flip ( dec2base (0:( p +1) ^ n - 1 , p +1) - '0 ' ,2) '
3 % Create an array of symbolic variables for the state
4 xsym = sym ( 'x ' ,[1 n ] , ' real ') ;
5 % Preallocate the matrix of symbolic variables
6 sym_univariate = sym ( ones ( size ( hpm ) ) ) ;
7 % Loop over the state variables to assign the polynomial
8 % according to the order and variable
9 for state_variable = 1 : n
10 sym_univariate ( state_variable ,1: end ) = hermiteH ( hpm (
state_variable ,1: end ) , xsym ( state_variable ) ) ;
11 end
12 base = prod ( sym_univariate ,1) ;
13 % The function omits the intercept ( first element ) .
14 % Otherwise , the evaluation of the whole training matrix
15 % is not possible at once , and the calculation should be
16 % achieved in a loop .
17 f = matlabFunction ( base (2: end ) , ' var ' ,{ xsym })

The function f can evaluate the complete set of training trajectories at once with the
omission of the first observable that corresponds to the intercept or constant value (the
consideration of this observable would require another programming strategy involving
loops, resulting in higher computational time and memory allocation). Notice the versatility
of using orthogonal polynomials, as the whole realm of available orthogonal polynomials
in Matlab is a valid choice, e.g., Laguerre, Legendre, Jacobi, etc. Note that the code snippet
Mathematics 2022, 10, 3859 6 of 18

defines the function of observables as a row vector, instead of the column vector notation
in the theoretical descriptions.
After the observables have been defined, their time evolution can be computed accord-
ing to
f ( x (k + 1)) = U f ( x (k)) + r ( x ), (6)
where U ∈ Rd×d is the matrix that provides the linear evolution of the observables and
r ( x ) is the error in the approximation. One of the main advantages of the EDMD algorithm
resides in the fact that the system description is linear in the function space. The solution
to (6) is the matrix U that minimizes the residual term r ( x ), which can be expressed as a
least-squares criterion:

N
1 1
l ( x, y) =
N ∑ 2 k f (yi ) − U f (xi )k22 , (7)
i =1

where N is the total number of samples in the training set. The ordinary least-squares (OLS)
solution is provided by
U = A/G, (8)
where the A/G notation replaces the inverse of the design matrix G, as even when using a
basis formed by the products of orthogonal polynomials, the design matrix can be close to
ill-conditioned (i.e., close to singular). This notation, particularly in Matlab, specifies that a
more robust algorithm compared to the inverse or pseudo-inverse is necessary to obtain
the approximation.
For the solution of (8), the matrices G, A ∈ Rd×d are defined by

1
G= f ( x ) f ( x )> (9)
N
1
A = f ( x ) f (y)> . (10)
N
Setting the observables as products of univariate orthogonal polynomials is an im-
provement, as it generally avoids the need to use a pseudo-inverse approach. Even though
the sequence of polynomials in the set of observables is no longer orthogonal, it is less likely
to have co-linear columns in the design matrix, improving the numerical stability of the so-
lution. With the training set and the observables, the method for calculating the regression
matrix U is shown in Listing 3. Notice that this code defines and uses all the arrays as their
transpose. This change is related to the approximation of the Koopman operator, where
it is necessary to calculate the right and left eigenvectors of U. The eigenfunctions of the
Koopman operator are determined from the left eigenvectors of the spectral decomposition.
In Matlab, the left eigenvectors result from additional algebraic manipulations of the right
eigenvectors and the diagonal matrix of eigenvalues, thereby decreasing the numerical
precision of the eigenfunctions. This problem is alleviated by computing U > and its spectral
decomposition so that the left eigenvectors are immediately available. In general, if U
is a normal matrix (diagonalizable), the additional steps involve the inverse of the right
eigenvectors to obtain the left eigenvectors and the calculation of this inverse, considering
again that the problem is close to being ill-conditioned, which reduces the accuracy of
the eigenfunctions.
Mathematics 2022, 10, 3859 7 of 18

Listing 3. Computation of the approximate Koopman operator for an OLS problem.

1 function u = getU (f , X , Y )
2 % GETU produces the U matrix of the decomposition
3
4 % Evaluate the snapshots with the f functions of the
5 % observables
6 x_eval = [ ones ( size (X ,1) ,1) , f ( X ) ];
7 y_eval = [ ones ( size (Y ,1) ,1) , f ( Y ) ];
8 % Calculates the G and A matrices
9 g = ( x_eval '* x_eval ) *(1/ size (X ,1) ) ;
10 a = ( x_eval '* y_eval ) *(1/ size (Y ,1) ) ;
11 % Notice that this returns the transpose
12 % of u
13 u = g\a;
14 end

Numerical Results with EDMD

Here, the EDMD algorithm is tested with the second case scenario for the Duffing
oscillator with hard damping. The EDMD algorithm can capture the dynamics of the
portion of the state space covered by the training set, which is therefore selected as the
outermost trajectory in Figure 2. The five remaining trajectories are used for testing. Table 1
provides the parameters of the original EDMD algorithm.

Table 1. Approximation parameters for the hard spring Duffing oscillator.

Parameter Value
α 1
β 1
δ 0.5
ics 4*rand(6,2)-2;
final_time 20
n_points 201
solver ode23s
tr_idex 3
ts_index [1 2 4 5 6]
polynomial laguerreL
N 199
p 4
observables 25
testing error 6.1370 × 10−05

In Figure 2, the graph on the left displays the phase plane of the system and shows the
training trajectory and testing trajectories along with their approximation by the EDMD
algorithm with the Laguerre polynomial basis. EDMD achieves a good approximation while
using only a small amount of data. However, notice that the discrete-time approximation
of a system of order 2 is of dimension twenty-five. In view of this dimensionality explosion
with regard to the original dimension of the state and the complexity of the system, it
is necessary to introduce reduction techniques that decrease the necessary number of
observables to increase the accuracy of the algorithm [24].
Mathematics 2022, 10, 3859 8 of 18

Figure 2. EDMD approximation of the hard damping Duffing oscillator.

2.2. pqEDMD Algorithm

The extension of the EDMD algorithm is a result of speculation on the good perfor-
mance of the original algorithm when coupled with a set of observables based on products
of univariate-orthogonal polynomials. The idea is to introduce a reduction method, based
on p-q quasi-norms, first introduced by Konakli and Sudret [27] for fault detection in poly-
nomial chaos problems. The reduction proceeds in the following way: if the q-quasi-norm
of the indexes that provide the order of the univariate-orthogonal polynomials is less than
the maximum order p of a particular observer, then this observer is eliminated from the set.
To implement this procedure, the orders of an observer are defined as αi , and the q-quasi
norm of these orders as
!1
n q
q
kαkq = ∑ αi , (11)
i =1

where q ∈ R+ and Equation (11) represent a norm only when q is an integer. When p is
redefined as the maximum order of a particular multivariate polynomial instead of the
maximum order of the univariate elements, the sets of polynomial orders that remain in
the basis are those that satisfy

α i = { α ∈ Nn : k α k q ≤ p } . (12)

The code snippet used to generate a set of observables based on Laguerre polynomials
with a maximum multivariate order p = 4 and a q-quasi-norm q = 0.7 is provided in
Listing 4.
The reduction of the basis is not only dimensional, as the p-q quasi-norm reduction
reduces the maximum order of the observables as well. As a rule of thumb (consider-
ing various application examples), the higher-order observables usually have a negative
contribution to the accuracy of the solution.
Mathematics 2022, 10, 3859 9 of 18

Listing 4. Generation of a p-q-reduced set of observables with a Laguerre basis.

1 p =4;
2 n =2;
3 q =0.7;
4 % Generate the matrix p of indexes
5 hpm = flip ( dec2base (0:( p +1) ^ n - 1 , p +1) - '0 ' ,2) ';
6 % Reduce the basis
7 orders = hpm (: , vecnorm ( hpm , q ) <= p ) ;
8 % Create an array of symbolic variables for the state
9 xsym = sym ( 'x ' ,[1 n ] , ' real ') ;
10 % Preallocate the matrix of symbolic variables
11 xPsi = sym ( ones ( size ( orders ) ) ) ;
12 % Loop over the state variables to assign the polynomial
13 % according to the order and variable
14 for state_variable = 1 : n
15 xPsi ( state_variable ,1: end ) = laguerreL ( orders (
state_variable ,1: end ) , xsym ( state_variable ) ) ;
16 end
17 base = prod ( xPsi ,1) ;
18 f = matlabFunction ( base (2: end ) , ' var ' ,{ xsym }) ;

Numerical Results with the pqEDMD

The algorithm is now applied to the Duffing oscillator with soft damping. The
pqEDMD algorithm can capture the dynamics of the two attractors provided that the
training set has at least one trajectory that converges to each of them. Additionally, as is the
case for the hard damping, each of these trajectories should be the outermost (see Figure 3).
Table 2 lists the parameters of the pqEDMD algorithm.

Table 2. Approximation parameters for the soft spring Duffing oscillator.

Parameter Value
α −1
β 1
δ 0.5
ics equal to hard spring
final_time 20
n_points 201
solver ode23s
tr_idex [1 3]
ts_index [2 4 5 6]
polynomial laguerreL
N 398
p 5
q ∞→1
observables 36 → 21
testing error 2.3770 × 10−4 → 1.1464 × 10−6

Even though the full basis achieves a low approximation error of 2.3770 × 10−4 , the re-
duction of the observables order reduces the empirical error by two orders of magnitude.
Comparing the dimension of the full basis to the reduced one does not represent a large
improvement. However, this result is due to the comparison between the best result after
performing a sweep over several p-q values. Imposing lower p-q values on the approxima-
tion has the potential to provide smaller sets of observables while sacrificing accuracy.
Mathematics 2022, 10, 3859 10 of 18

Figure 3. pqEDMD approximation for the soft damping Duffing oscillator.

Next, the first case scenario is considered; here, the damping parameter is zero and
the system oscillates around the fixed point at the origin. The pqEDMD algorithm can
capture the dynamics of the system if the innermost and outermost limit cycles compose
the training set; otherwise, the algorithm cannot capture the dynamics. Table 3 shows a
summary of the simulations along with the results; it is apparent that even though the
empirical error is higher than in the other two case scenarios, the approximation is accurate
(see Figure 4).

Table 3. Approximation parameters for the undamped Duffing oscillator.

Parameter Value
α 1
β 1
δ 0
ics same as hard spring
final_time 20
n_points 201
solver ode23s
tr_idex [3 4]
ts_index [1 2 5 6]
polynomial hermiteH
N 398
p 4
q ∞ → 1.1
observables 25 → 15
testing error 0.1209 → 0.0023

The sweep over different p-q values provides a reduced basis with lower error than
with the full basis.
Mathematics 2022, 10, 3859 11 of 18

Figure 4. pqEDMD approximation for the undamped Duffing oscillator.

Even though p-q quasi-norm reduction produces more accurate and tractable solutions,
having products of orthogonal univariate polynomials does not necessarily produce an
orthogonal basis. In certain scenarios, the evaluation of the observables produces an ill-
conditioned design matrix G. Therefore, the next section proposes a way to eliminate even
more observables from the basis, improving the numerical stability of the solution.

2.3. Improving Numerical Stability via QR Decomposition

QR decomposition [28] can be used to improve the numerical stability and reduce
the number of observables even further. If we assume that the design matrix G ∈ Rd×d in
Equation (9) is obtained based on the products of orthogonal polynomials and that there
are no co-linear columns, or, in other words, that rank( G ) = d holds, then it is possible to
decompose this matrix into the product

G = QR, (13)

where Q ∈ Rd×d is orthogonal, i.e., Q> Q = Id and R ∈ Rd×d is upper triangular. Column
pivoting methods for QR decomposition rely on exchanging the rows of G such that in
every step of the diagonalization of R and the subsequent calculation of the orthogonal
columns of Q the procedure starts with a column that is as independent as possible from
the columns of G already processed. This method yields a permutation matrix P ∈ Rd×d
such that
GP = QR, (14)
where the permutation of columns makes the absolute value of the diagonal elements
in R non-increasing, i.e., |r1,1 | ≥ |r2,2 | ≥ · · · ≥ |rd,d |. Furthermore, considering that the
permutation process selects the most linearly independent column of G in every step of
the process, the last columns in the analysis are the ones that are close to being co-linear.
Therefore, eliminating the observable related to the last column improves the residual
condition number of G. The modified function for the calculation of the regression matrix
U is provided in Listing 5.
Mathematics 2022, 10, 3859 12 of 18

Listing 5. Computation of the regression matrix based on QR decomposition.

1 function u = getU ( obj , X , Y )

2 % GETU produces the U matrix of the decomposition
3
4 % Evaluate the snapshots with the f function of the
5 % observables ( called psi in the following )
6 Psi = obj . VVFO . Psi ;
7 x_eval = [ ones ( size (X ,1) ,1) , Psi ( X ) ];
8 y_eval = [ ones ( size (Y ,1) ,1) , Psi ( Y ) ];
9 % Calculates the G matrix
10 g = ( x_eval '* x_eval ) *(1/ size (X ,1) ) ;
11 % qr decomposition and elimination
12 % of near - co - linear observables
13 while rcond ( g ) <= eps
14 [~ ,~ , E ] = qr (g ,0) ;
15 % Check that the last element is not an
16 % order one polynomial , if it is ,
17 % shift the array
18 while any ( E ( end ) == find ( sum ( obj . VVFO . polyn omials_or der )
==1) )
19 E = circshift (E ,1) ;
20 end
21 % Eliminate the observable correspondint to the last
element
22 % of E
23 obj . VVFO . poly nomials_ order = obj . VVFO . polynomi als_order
(: , sort ( E (1: end -1) ) ) ;
24 % Get the observables matlabFunction updated
25 Psi = obj . VVFO . Psi ;
26 % Evaluate again
27 if all (~ logical ( obj . VVFO . polyn omials_o rder (: ,1) ) )
28 x_eval = [ ones ( size (X ,1) ,1) , Psi ( X ) ];
29 y_eval = [ ones ( size (Y ,1) ,1) , Psi ( Y ) ];
30 else
31 x_eval = Psi ( X ) ;
32 y_eval = Psi ( Y ) ;
33 end
34 g = ( x_eval '* x_eval ) *(1/ size (X ,1) ) ;
35 end
36 a = ( x_eval '* y_eval ) *(1/ size (Y ,1) ) ;
37
38 % This returns the transpose of U because it avoids
39 % numerical errors if it is considered as a Koopman
operator
40 % and the espectral decomposition is necessary
41 u = g\a;
42 end

In addition, the code snippet shows particular aspects of the overall solution. First,
an object containing the observables, i.e., the matlabFunction obj.VVFO.Psi, replaces
the original matlabFunction f for the evaluation of the snapshots. Second, note that the
exclusion of observables avoids the elimination of the first order univariate polynomials in
the basis, as they are used to recover the state. Finally, the method checks for the existence
of the constant observable or the intercept, as it could be eliminated due to being close to
co-linear with another observable, which we obviously do not want to happen.
Mathematics 2022, 10, 3859 13 of 18

2.4. Matlab Package

pqEDMD() is the main class of the Matlab package that provides an array of decom-
positions based on the pqEDMD algorithm pqEDMD_array. The cardinality of the array of
solutions may be less than the product of the cardinality of p and q, as certain p-q pairs
produce the same set of indices, i.e., the algorithm would compute the same decomposition
more than once. In addition, it calculates the empirical error of the approximations based
on the test set and returns the best-performing approximation from the array as a separate
attribute best_pqEDMD. The code provides the complete set of solutions, as a user may opt
to use a compact solution that is not as accurate as the best one for tractability reasons,
e.g., an MPC controller, where a smaller basis guarantees feasibility for longer horizons
and has a lower computational cost.
The only required input for pqEDMD() is the system argument, where it is necessary to
provide a structure array with the Time and SV fields with at least two trajectories in the
array, one for training and one for testing. The remaining arguments are optional, e.g., the
array of positive-integer values p, the array of positive values q, the structure of training
and testing trajectories with the fields tr_index and ts_index, the string specifying the
type of polynomial, the array of polynomial parameters (if the polynomial type is either
“Jacobi” or “Gegenbauer”), the boolean flag of normalization, and the string indicating
the decomposition method. For example, Listing 6 shows a call to the algorithm with a
complete set of arguments.

Listing 6. Complete call to the pqEDMD algorithm.

1 pqEDMD ( system , [3 4 5] , [0.2 0.5 0.7 inf ] ,...

2 ' polynomial ' , ' Jacobi ' , ' polyParam ' ,[2 3] , ' method ' , ' OLS
' ,...
3 ' tr_ts ' , struct ( ' tr_index ' , tr_index , ' ts_index ' ,
ts_index ) ,...
4 ' normalization ' , false ) ;

To provide the different approximations in the main class, pqVVFO() handles the
observables for different values of p, q, and the polynomial type. Its output is the matrix of
polynomial indexes, a symbolic vector of observable functions, and a matlabFunction Psi
to evaluate the observables arithmetically and efficiently; it accepts a matrix of values,
avoiding evaluation with loops.
The remaining classes are the implementations of different decompositions based
on different algorithms. The ExtendedDecomposition() is the traditional least-squares
method described in this article. In addition, there are two additional available decompo-
sitions. MaxLikeDecomposition() is used for data with noise, where the maximum likeli-
hood algorithm assumes that the transformation of the states in the function space preserves
a unimodal Gaussian distribution of the noise in the state space (this is a work in progress;
preliminary results can be found in [29]). These properties of the distribution of noise in the
function space are a strong assumption; nonetheless, it is sometimes possible to identify dy-
namical systems corrupted with noise. The last decomposition leverages the advantages of
regularized lasso regression to produce sparse solutions, i.e., RegularizedDecomposition().
Even though the solutions are more tractable, the regularized method sacrifices accuracy.
Figure 5 shows the architecture of the solution with the relationship between classes.
The current functionality requires the user to call pqEDMD() with the appropriate inputs
and options in order to obtain an array of decompositions. This class handles the creation
of the necessary pqVVFO() objects to feed into the required decomposition. It is possible to
use and extend the observable class to use in other types of decompositions without the
use of the main class. The code is available for download in the Supplementary Materials
section of this paper.
 1, 03859
Mathematics 2022, 10, 14 of 18
Mathematics 2022, 1, 0 14 of 18

Figure 5. Package architecture.

Figure 5. Package architecture.
3. An Additional Application Example
3. An Additional Application Example
To conclude this article, an additional case study is discussed involving a set of
To conclude this article, an additional case study is discussed involving a set of
reactions occurring in a continuously stirred tank reactor, described by
reactions occurring in a continuously stirred tank reactor, described by
r1
s1 +2s3 r1 s2 +3s3
s1 +2s3 s2 +3s3
r2
s2 +2s4 r2 3s4 (15)
s2 +2s4 3s4
r3
s3 r3 s5
s3 s5
r4
s4 r4 s5
s4 s5 (15)
(15)
where there are two types of substrate, s11 and s22. The first substrate is the only component
where there are two types of substrate, s and s2 . The first substrate is the only component
in the inflow, and is the only component1necessary for the replication of the first species s3
in the inflow, and is the only component necessary for the replication of the first species s3
according to the replication rate constant r11. In addition to the replication of s33, the product3
according to the replication rate constant r . In addition to the replication of s3 , the product
of the first reaction is the second substrate1s22, which in turn is necessary for the replication
of the first reaction is the second substrate s2 , which in turn is necessary for the replication
of the second species s44 according to the replication rate constant r22. The remaining variable
of the second species s according to the replication rate constant r . The remaining variable
s55 is the combination 4of the dead species from the two groups,2 where each group dies
s5 is the combination of the dead species from the two groups, where each group dies
according to the reaction rates r3 and r4 , respectively. The ordinary differential equations
according to the reaction rates r33 and r44 , respectively. The ordinary differential equations
Mathematics 2022, 10, 3859 15 of 18

that describe the dynamics of network (15) according to the polynomial formulation of
mass action kinetics [30] and the material exchange with the environment are provided by

ẋ1 = −r1 x1 x32 + d − dx1

ẋ2 = +r1 x1 x32 − r2 x2 x42 − dx2
ẋ3 = +r1 x1 x32 − r3 x3 − dx3 (16)
ẋ4 = +r2 x2 x42
− r4 x4 − dx4
ẋ5 = +r3 x3 + r4 x4 − dx5 ,

where d = 0.5 is the in/out-flow (dilution rate) of the system and the values for the reaction
rates are r = [7 5 0.3 0.05]> . With these rate constants, the system has three asymptotically
stable points: the working point, where the two species s3 and s4 coexist, a point where
species s3 thrives and species s4 washes out, and a wash-out point, where the concentration
of both species vanishes. To construct the database, the strategy is to generate a set of orbits
with an even distribution of initial conditions converging to each of the equilibrium points.
Certain trajectories converging to each point are used as the training set to produce a linear
expanded approximation of the system.
The set of orbits is taken from the numerical integration of the ODE (16) via the ode23s
method with an output sampling ∆t = 0.1 for an arbitrary number of initial conditions
until the full set of orbits has a total of 20 trajectories that converge to each point, resulting
in 60 trajectories in total for the execution of the algorithm.
From each of the sets of orbits that converge to the fixed points, 50% are used for
the approximation and the remaining for testing the solution. It is important to have a
training set with sufficient information about the trajectories of the system, and in a similar
way as for the second and third scenarios of the Duffing equation, to select the trajectories
that are far away from the equilibrium point. For system (16), the choice of trajectories
for the training set are the ten trajectories that at any given time are furthest away from
the equilibrium point to which they converge. Figure 6 shows a selection of training and
testing trajectories.

tr
ts

Figure 6. Training and testing trajectories of the biochemical reaction system.

Assuming that the orbits of the system are in a structure array with the appropriate
fields named system and that the training and testing indexes for the approximation have
Mathematics 2022, 10, 3859 16 of 18

been carefully selected and placed in a structure named tr_ts, the call to the pqEDMD()
class that provides an accurate approximation is shown in Listing 7, where the solution is
obtained through the default decomposition method (ordinary least squares), the default
polynomial type (Legendre), and without normalization.

Listing 7. Class call to approximate the reaction network dynamics.

1 % the pqEDMD !!!

2 p = [3 4 5];
3 q = [0.5 0.7 0.9];
4 mak_net_approx = pqEDMD ( system ,p ,q , ' tr_ts ' , tr_ts ) ;

Table 4 shows a summary of the simulation parameters used to generate the orbits
and to obtain the approximation. The clear advantage of using the reduction method lies
in the comparison between the full basis of polynomials, i.e., from 3125 observables for
p = 4 and a system of five state variables to a basis of 51 polynomials for q = 0.7. Although
the computation with a full basis is computationally intensive, it leads to a solution that is
not satisfactory, as the state matrix is not Hurwitz and the trajectories diverge, leading to a
result with an infinite error metric.
Figure 7 depicts the comparison of several testing trajectories with their corresponding
approximations. It is clear that certain trajectories converge to a different fixed point than
the one they are supposed to. This phenomenon causes the empirical error grow while
remaining bounded. The reason for this behavior is the lack of training trajectories near
the boundary of the attraction regions of the asymptotically stable equilibrium points. For
better performance of the algorithm in terms of the number of orbits necessary for the
approximation, and possibly the dimension of the observable basis, an experimental design
procedure is required.

ts
approx

Figure 7. Comparison between the testing and approximation trajectories of the biochemical reac-
tion system.
Mathematics 2022, 10, 3859 17 of 18

Table 4. Approximation parameters for the reaction network.

Parameter Value
r1 7
r2 5
r3 0.3
r4 0.05
ics 20 per x ∗
final_time stoped when converges
delta_t 0.1
solver ode15s
tr_idex max(vecnorm(SV-x_ast))
ts_index ~tr_index
polynomial Laguerre
N 3874
p 4
q ∞ → 0.7
observables 3125 → 51
testing error ∞ → 12.94

4. Conclusions
This paper presents a methodology to derive discrete-time approximations of nonlin-
ear dynamical systems via the pqEDMD algorithm and proposes a Matlab library that can
hopefully help popularize the use of the method by non-expert users. The discussion of
the methodology and codes is illustrated with several case studies related to the Duffing
oscillator and by an example involving a biochemical reaction network.

Supplementary Materials: The following supporting information can be downloaded at: https:
//www.mdpi.com/article/10.3390/math10203859/s1. A Matlab library is readily available in the
Supplementary Materials.
Author Contributions: Conceptualization, C.G.-T. and A.V.W.; methodology, C.G.-T. and A.V.W.;
software, C.G.-T.; writing—review and editing, C.G.-T. and A.V.W. All authors have read and agreed
to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Williams, M.O.; Kevrekidis, I.G.; Rowley, C.W. A Data-Driven Approximation of the Koopman Operator: Extending Dynamic
Mode Decomposition. J. Nonlinear Sci. 2015, 25, 1307–1346. [CrossRef]
2. Ljung, L. System Identification: Theory for the User, 2nd ed.; Prentice Hall: Hoboken, NJ, USA, 1999.
3. John, L.; Crassidis, J.L.J. Optimal Estimation of Dynamic Systems, 1st ed.; Chapman & Hall/CRC Applied Mathematics & Nonlinear
Science; Chapman and Hall/CRC: Boca Raton, FL, USA, 2004.
4. Penny, W.; Harrison, L. CHAPTER 40—Multivariate autoregressive models. In Statistical Parametric Mapping; Friston, K.,
Ashburner, J., Kiebel, S., Nichols, T., Penny, W., Eds.; Academic Press: London, UK, 2007; pp. 534–540.
5. Roger, G.; Ghanem, P.D.S. Stochastic Finite Elements: A Spectral Approach, Revised ed.; Dover Publications: Mineola, NY, USA,
2003.
6. Stewart, G.W. On the Early History of the Singular Value Decomposition. SIAM Rev. 1993, 35, 551–566. [CrossRef]
7. Chatterjee, A. An introduction to the proper orthogonal decomposition. Curr. Sci. 2000, 78, 808–817.
8. Schmid, P.J. Dynamic mode decomposition of numerical and experimental data. J. Fluid Mech. 2010, 656, 5–28. [CrossRef]
9. Mezić, I. Spectral Properties of Dynamical Systems, Model Reduction and Decompositions. Nonlinear Dyn. 2005, 41, 309–325.
[CrossRef]
10. Mezić, I.; Mezi, I. On applications of the spectral theory of the Koopman operator in dynamical systems and control theory. In
Proceedings of the 2015 54th IEEE Conference on Decision and Control (CDC), Osaka, Japan, 16–18 December 2015; pp. 7034–7041.
11. Mezic, I.; Surana, A. Koopman Mode Decomposition for Periodic/Quasi-periodic Time Dependence—The funding provided by
UTRC is greatly appreciated. IFAC-PapersOnLine 2016, 49, 690–697. [CrossRef]
Mathematics 2022, 10, 3859 18 of 18

12. Koopman, B.O. Hamiltonian Systems and Transformation in Hilbert Space. Proc. Natl. Acad. Sci. USA 1931, 17, 315–318.
[CrossRef]
13. Mauroy, A.; Mezić, I. A spectral operator-theoretic framework for global stability. In Proceedings of the 52nd IEEE Conference on
Decision and Control, Firenze, Italy, 10–13 December 2013; pp. 5234–5239.
14. Mauroy, A.; Mezic, I. Global Stability Analysis Using the Eigenfunctions of the Koopman Operator. IEEE Trans. Autom. Control
2016, 61, 3356–3369. [CrossRef]
15. Garcia-Tenorio, C.; Tellez-Castro, D.; Mojica-Nava, E.; Vande Wouwer, A. Evaluation of the Region of Attractions of Higher
Dimensional Hyperbolic Systems using the Extended Dynamic Mode Decomposition. arXiv 2022, arXiv:2209.02028.
16. Brunton, S.L.; Brunton, B.W.; Proctor, J.L.; Kutz, J.N. Koopman invariant subspaces and finite linear representations of nonlinear
dynamical systems for control. PLoS ONE 2016, 11, e0150171. [CrossRef]
17. Korda, M.; Mezić, I. Linear predictors for nonlinear dynamical systems: Koopman operator meets model predictive control.
Automatica 2018, 93, 149–160. [CrossRef]
18. Tellez-Castro, D.; Garcia-Tenorio, C.; Mojica-Nava, E.; Sofrony, J.; Vande Wouwer, A. Data-Driven Predictive Control of
Interconnected Systems Using the Koopman Operator. Actuators 2022, 11, 151. [CrossRef]
19. Garcia-Tenorio, C.; Vande Wouwer, A. Extended Predictive Control of Interconnected Oscillators. In Proceedings of the 2022
8th International Conference on Control, Decision and Information Technologies (CoDIT), Istanbul, Turkey, 17–20 May 2022;
Volume 1, pp. 15–20.
20. Williams, M.O.; Rowley, C.W.; Kevrekidis, I.G. A kernel-based method for data-driven koopman spectral analysis. J. Comput.
Dyn. 2016, 2, 247–265. [CrossRef]
21. Kaiser, E.; Kutz, J.N.; Brunton, S.L. Sparse identification of nonlinear dynamics for model predictive control in the low-data limit.
Proc. R. Soc. A Math. Phys. Eng. Sci. 2018, 474, 20180335. [CrossRef] [PubMed]
22. Li, Q.; Dietrich, F.; Bollt, E.M.; Kevrekidis, I.G. Extended dynamic mode decomposition with dictionary learning: A data-driven
adaptive spectral decomposition of the Koopman operator. Chaos Interdiscip. J. Nonlinear Sci. 2017, 27, 103111. [CrossRef]
23. Proctor, J.L.; Brunton, S.L.; Kutz, J.N. Dynamic Mode Decomposition with Control. SIAM J. Appl. Dyn. Syst. 2016, 15, 142–161.
[CrossRef]
24. Garcia-Tenorio, C.; Delansnay, G.; Mojica-Nava, E.; Vande Wouwer, A. Trigonometric Embeddings in Polynomial Extended Mode
Decomposition—Experimental Application to an Inverted Pendulum. Mathematics 2021, 9, 1119. [CrossRef]
25. Garcia-Tenorio, C.; Mojica-Nava, E.; Sbarciog, M.; Vande Wouwer, A. Analysis of the ROA of an anaerobic digestion process via
data-driven Koopman operator. Nonlinear Eng. 2021, 10, 109–131. [CrossRef]
26. Takata, H. Transformation of a nonlinear system into an augmented linear system. IEEE Trans. Autom. Control 1979, 24, 736–741.
[CrossRef]
27. Konakli, K.; Sudret, B. Polynomial meta-models with canonical low-rank approximations: Numerical insights and comparison to
sparse polynomial chaos expansions. J. Comput. Phys. 2016, 321, 1144–1169. [CrossRef]
28. Gander, W. Algorithms for the QR decomposition. Res. Rep 1980, 80, 1251–1268.
29. Garcia-Tenorio, C.; Vande Wouwer, A. Maximum Likelihood pqEDMD Identification. In Proceedings of the 26th International
Conference on System Theory, Control and Computing 2022 (ICSTCC), Sinaia, Romania, 19–21 October 2022.
30. Chellaboina, V.; Bhat, S.P.; Haddad, W.M.; Bernstein, D.S. Modeling and analysis of mass-action kinetics. IEEE Control Syst. Mag.
2009, 29, 60–78.