-
Deep Learning for Structure-Preserving Universal Stable Koopman-Inspired Embeddings for Nonlinear Canonical Hamiltonian Dynamics
Authors:
Pawan Goyal,
Süleyman Yıldız,
Peter Benner
Abstract:
Discovering a suitable coordinate transformation for nonlinear systems enables the construction of simpler models, facilitating prediction, control, and optimization for complex nonlinear systems. To that end, Koopman operator theory offers a framework for global linearization for nonlinear systems, thereby allowing the usage of linear tools for design studies. In this work, we focus on the identi…
▽ More
Discovering a suitable coordinate transformation for nonlinear systems enables the construction of simpler models, facilitating prediction, control, and optimization for complex nonlinear systems. To that end, Koopman operator theory offers a framework for global linearization for nonlinear systems, thereby allowing the usage of linear tools for design studies. In this work, we focus on the identification of global linearized embeddings for canonical nonlinear Hamiltonian systems through a symplectic transformation. While this task is often challenging, we leverage the power of deep learning to discover the desired embeddings. Furthermore, to overcome the shortcomings of Koopman operators for systems with continuous spectra, we apply the lifting principle and learn global cubicized embeddings. Additionally, a key emphasis is paid to enforce the bounded stability for the dynamics of the discovered embeddings. We demonstrate the capabilities of deep learning in acquiring compact symplectic coordinate transformation and the corresponding simple dynamical models, fostering data-driven learning of nonlinear canonical Hamiltonian systems, even those with continuous spectra.
△ Less
Submitted 26 August, 2023;
originally announced August 2023.
-
Linearly Implicit Global Energy Preserving Reduced-order Models for Cubic Hamiltonian Systems
Authors:
Süleyman Yildiz,
Pawan Goyal,
Peter Benner
Abstract:
This work discusses the model reduction problem for large-scale multi-symplectic PDEs with cubic invariants. For this, we present a linearly implicit global energy-preserving method to construct reduced-order models. This allows to construct reduced-order models in the form of Hamiltonian systems suitable for long-time integration. Furthermore, We prove that the constructed reduced-order models pr…
▽ More
This work discusses the model reduction problem for large-scale multi-symplectic PDEs with cubic invariants. For this, we present a linearly implicit global energy-preserving method to construct reduced-order models. This allows to construct reduced-order models in the form of Hamiltonian systems suitable for long-time integration. Furthermore, We prove that the constructed reduced-order models preserve global energy, and the spatially discrete equations also preserve the spatially-discrete local energy conversation law. We illustrate the efficiency of the proposed method using three numerical examples, namely a linear wave equation, the Korteweg-de Vries equation, and the Camassa-Holm equation, and present a comparison with the classical POD-Galerkin method.
△ Less
Submitted 4 August, 2023;
originally announced August 2023.
-
Analytical Benchmark Problems for Multifidelity Optimization Methods
Authors:
L. Mainini,
A. Serani,
M. P. Rumpfkeil,
E. Minisci,
D. Quagliarella,
H. Pehlivan,
S. Yildiz,
S. Ficini,
R. Pellegrini,
F. Di Fiore,
D. Bryson,
M. Nikbay,
M. Diez,
P. Beran
Abstract:
The paper presents a collection of analytical benchmark problems specifically selected to provide a set of stress tests for the assessment of multifidelity optimization methods. In addition, the paper discusses a comprehensive ensemble of metrics and criteria recommended for the rigorous and meaningful assessment of the performance of multifidelity strategies and algorithms.
The paper presents a collection of analytical benchmark problems specifically selected to provide a set of stress tests for the assessment of multifidelity optimization methods. In addition, the paper discusses a comprehensive ensemble of metrics and criteria recommended for the rigorous and meaningful assessment of the performance of multifidelity strategies and algorithms.
△ Less
Submitted 16 April, 2022;
originally announced April 2022.
-
Intrusive and non-intrusive reduced order modeling of the rotating thermal shallow water equation
Authors:
Süleyman Yıldız,
Murat Uzunca,
Bülent Karasözen
Abstract:
In this paper, we investigate projection-based intrusive and data-driven non-intrusive model order reduction methods in numerical simulation of rotating thermal shallow water equation (RTSWE) in parametric and non-parametric form. Discretization of the RTSWE in space with centered finite differences leads to Hamiltonian system of ordinary differential equations with linear and quadratic terms. The…
▽ More
In this paper, we investigate projection-based intrusive and data-driven non-intrusive model order reduction methods in numerical simulation of rotating thermal shallow water equation (RTSWE) in parametric and non-parametric form. Discretization of the RTSWE in space with centered finite differences leads to Hamiltonian system of ordinary differential equations with linear and quadratic terms. The full-order model (FOM) is obtained by applying linearly implicit Kahan's method in time. Applying proper orthogonal decomposition with Galerkin projection (POD-G), we construct the intrusive reduced-order model (ROM). We apply operator inference (OpInf) with re-projection for non-intrusive reduced-order modeling. In the parametric case, we make use of the parameter dependency at the level of the PDE without interpolating between the reduced operators. The least-squares problem of the OpInf is regularized with the minimum norm solution. Both ROMs behave similar and are able to accurately predict the test and training data and capture system behavior in the prediction phase with several orders of computational speedup over the FOM. The preservation of system physics such as the conserved quantities of the RTSWE by both ROMs enables that the models fit better to data and stable solutions are obtained in long-term predictions, which are robust to parameter changes.
△ Less
Submitted 14 April, 2021; v1 submitted 31 March, 2021;
originally announced April 2021.
-
alfonso: Matlab package for nonsymmetric conic optimization
Authors:
Dávid Papp,
Sercan Yıldız
Abstract:
We present alfonso, an open-source Matlab package for solving conic optimization problems over nonsymmetric convex cones. The implementation is based on the authors' corrected analysis of a primal-dual interior-point method of Skajaa and Ye. This method enables optimization over any convex cone as long as a logarithmically homogeneous self-concordant barrier is available for the cone or its dual.…
▽ More
We present alfonso, an open-source Matlab package for solving conic optimization problems over nonsymmetric convex cones. The implementation is based on the authors' corrected analysis of a primal-dual interior-point method of Skajaa and Ye. This method enables optimization over any convex cone as long as a logarithmically homogeneous self-concordant barrier is available for the cone or its dual. This includes many nonsymmetric cones, for example, hyperbolicity cones and their duals (such as sum-of-squares cones), semidefinite and second-order cone representable cones, power cones, and the exponential cone. Besides enabling the solution of problems which cannot be cast as optimization problems over a symmetric cone, it also offers performance advantages for problems whose symmetric cone programming representation requires a large number of auxiliary variables or has a special structure that can be exploited in the barrier computation.
The worst-case iteration complexity of alfonso is the best known for non-symmetric cone optimization: $O(\sqrtν\log(1/ε))$ iterations to reach an $ε$-optimal solution, where $ν$ is the barrier parameter of the barrier function used in the optimization.
alfonso can be interfaced with a Matlab function (supplied by the user) that computes the Hessian of a barrier function for the cone. For convenience, a simplified interface is also available to optimize over the direct product of cones for which a barrier function has already been built into the software. This interface can be easily extended to include new cones. Both interfaces are illustrated by solving linear programs. The oracle interface and the efficiency of alfonso are also demonstrated using a design of experiments problem in which the tailored barrier computation greatly decreases the solution time compared to using state-of-the-art conic optimization software.
△ Less
Submitted 11 January, 2021;
originally announced January 2021.
-
Energy preserving reduced-order modelling of thermal shallow water equation
Authors:
Suleyman Yildiz,
Murat Uzunca,
Bulent Karasozen
Abstract:
In this paper, Hamiltonian and energy preserving reduced-order models are developed for the rotating thermal shallow water equation (RTSWE) in the non-canonical Hamiltonian form with the state-dependent Poisson matrix. The high fidelity full solutions are obtained by discretizing the RTSWE in space with skew-symmetric finite-differences, that preserve the Hamiltonian structure. The resulting skew-…
▽ More
In this paper, Hamiltonian and energy preserving reduced-order models are developed for the rotating thermal shallow water equation (RTSWE) in the non-canonical Hamiltonian form with the state-dependent Poisson matrix. The high fidelity full solutions are obtained by discretizing the RTSWE in space with skew-symmetric finite-differences, that preserve the Hamiltonian structure. The resulting skew-gradient system is integrated in time with the energy preserving average vector field (AVF) method. The reduced-order model (ROM) is constructed in the same way as the full order model (FOM), preserving the reduced skew-symmetric structure and integrating in time with the AVF method. Relying on structure-preserving discretizations in space and time and applying proper orthogonal decomposition (POD) with the Galerkin projection, an energy preserving reduced order model (ROM) is constructed. The nonlinearities in the ROM are computed by applying the discrete empirical interpolation (DEIM) method to reduce the computational cost. The computation of the reduced-order solutions is accelerated further by the use of tensor techniques. The overall procedure yields a clear separation of the offline and online computational cost of the reduced solutions. The accuracy and computational efficiency of the ROMs are demonstrated for a numerical test problem. Preservation of the energy (Hamiltonian), and other conserved quantities, i.e. mass, buoyancy, and total vorticity show that the reduced-order solutions ensure the long-term stability of the solutions while exhibiting several orders of magnitude computational speedup over the FOM.
△ Less
Submitted 16 February, 2021; v1 submitted 1 November, 2020;
originally announced November 2020.
-
Data-Driven Learning of Reduced-order Dynamics for a Parametrized Shallow Water Equation
Authors:
Süleyman Yıldız,
Pawan Goyal,
Peter Benner,
Bülent Karasözen
Abstract:
This paper discusses a non-intrusive data-driven model order reduction method that learns low-dimensional dynamical models for a parametrized shallow water equation. We consider the shallow water equation in non-traditional form (NTSWE). We focus on learning low-dimensional models in a non-intrusive way. That means, we assume not to have access to a discretized form of the NTSWE in any form. Inste…
▽ More
This paper discusses a non-intrusive data-driven model order reduction method that learns low-dimensional dynamical models for a parametrized shallow water equation. We consider the shallow water equation in non-traditional form (NTSWE). We focus on learning low-dimensional models in a non-intrusive way. That means, we assume not to have access to a discretized form of the NTSWE in any form. Instead, we have snapshots that are obtained using a black-box solver. Consequently, we aim at learning reduced-order models only from the snapshots. Precisely, a reduced-order model is learnt by solving an appropriate least-squares optimization problem in a low-dimensional subspace. Furthermore, we discuss computational challenges that particularly arise from the optimization problem being ill-conditioned. Moreover, we extend the non-intrusive model order reduction framework to a parametric case where we make use of the parameter dependency at the level of the partial differential equation. We illustrate the efficiency of the proposed non-intrusive method to construct reduced-order models for NTSWE and compare it with an intrusive method (proper orthogonal decomposition). We furthermore discuss the predictive capabilities of both models outside the range of the training data.
△ Less
Submitted 4 August, 2020; v1 submitted 28 July, 2020;
originally announced July 2020.
-
Reduced order modelling of nonlinear cross-diffusion systems
Authors:
Bülent Karasözen,
Gülden Mülayim,
Murat Uzunca,
Süleyman Yıldız
Abstract:
In this work, we present a reduced-order model for a nonlinear cross-diffusion problem from population dynamics, for the Shigesada-Kawasaki-Teramoto (SKT) equation with Lotka-Volterra kinetics. The finite-difference discretization of the SKT equation in space results in a system of linear--quadratic ordinary differential equations (ODEs). The reduced order model (ROM) has the same linear-quadratic…
▽ More
In this work, we present a reduced-order model for a nonlinear cross-diffusion problem from population dynamics, for the Shigesada-Kawasaki-Teramoto (SKT) equation with Lotka-Volterra kinetics. The finite-difference discretization of the SKT equation in space results in a system of linear--quadratic ordinary differential equations (ODEs). The reduced order model (ROM) has the same linear-quadratic structure as the full order model (FOM). Using the linear-quadratic structure of the ROM, the reduced-order solutions are computed independent of the full solutions with the proper orthogonal decomposition (POD). The computation of the reduced solutions is further accelerated by applying tensorial POD. The formation of the patterns of the SKT equation consists of a fast transient phase and a long steady-state phase. Reduced order solutions are computed by separating the time, into two-time intervals. In numerical experiments, we show for one-and two-dimensional SKT equations with pattern formation, the reduced-order solutions obtained in the time-windowed form, i.e., principal decomposition framework (P-POD), are more accurate than the global POD solutions (G-POD) obtained in the whole time interval. Furthermore, we show the decrease of the entropy numerically by the reduced solutions, which is important for the global existence of nonlinear cross-diffusion equations such as the SKT equation.
△ Less
Submitted 8 February, 2021; v1 submitted 9 June, 2020;
originally announced June 2020.
-
Structure-preserving reduced-order modelling of Korteweg de Vries equation
Authors:
Bulent Karasozen,
Murat Uzunca,
Suleyman Yildiz
Abstract:
Computationally efficient, structure-preserving reduced-order methods are developed for the Korteweg de Vries (KdV) equations in Hamiltonian form. The KdV equation is discretized in space by finite differences. The resulting skew-gradient system of ordinary differential equations (ODEs) is integrated with the linearly implicit Kahan's method, which preserves the Hamiltonian approximately. We have…
▽ More
Computationally efficient, structure-preserving reduced-order methods are developed for the Korteweg de Vries (KdV) equations in Hamiltonian form. The KdV equation is discretized in space by finite differences. The resulting skew-gradient system of ordinary differential equations (ODEs) is integrated with the linearly implicit Kahan's method, which preserves the Hamiltonian approximately. We have shown, using proper orthogonal decomposition (POD), the Hamiltonian structure of the full-order model (FOM) is preserved by the reduced-order model (ROM). The quadratic nonlinear terms of the KdV equation are evaluated efficiently by the use of tensorial methods, clearly separating the offline-online cost of the FOMs and ROMs. The accuracy of the reduced solutions, preservation of the Hamiltonian, momentum and mass, and computational speed-up gained by ROMs are demonstrated for the one-dimensional KdV equation, coupled KdV equations and two-dimensional Zakharov-Kuznetzov equation with soliton solutions
△ Less
Submitted 7 February, 2021; v1 submitted 17 April, 2020;
originally announced April 2020.
-
Structure-preserving Reduced Order Modeling of non-traditional Shallow Water Equation
Authors:
Süleyman Yıldız,
Murat Uzunca,
Bülent Karasözen
Abstract:
An energy preserving reduced order model is developed for the nontraditional shallow water equation (NTSWE) with full Coriolis force. The NTSWE in the noncanonical Hamiltonian/Poisson form is discretized in space by finite differences. The resulting system of ordinary differential equations is integrated in time by the energy preserving average vector field (AVF) method. The Poisson structure of t…
▽ More
An energy preserving reduced order model is developed for the nontraditional shallow water equation (NTSWE) with full Coriolis force. The NTSWE in the noncanonical Hamiltonian/Poisson form is discretized in space by finite differences. The resulting system of ordinary differential equations is integrated in time by the energy preserving average vector field (AVF) method. The Poisson structure of the NTSWE in discretized exhibits a skew-symmetric matrix depending on the state variables. An energy preserving, computationally efficient reduced-order model (ROM) is constructed by proper orthogonal decomposition with Galerkin projection. The nonlinearities are computed for the ROM efficiently by discrete empirical interpolation method. Preservation of the semi-discrete energy and the enstrophy are shown for the full order model, and for the ROM which ensures the long term stability of the solutions. The accuracy and computational efficiency of the ROMs are shown by two numerical test problems
△ Less
Submitted 26 February, 2020;
originally announced February 2020.
-
Structure Preserving Model Order Reduction of Shallow Water Equations
Authors:
Bülent Karasözen,
Süleyman Yıldız,
Murat Uzunca
Abstract:
In this paper, we present two different approaches for constructing reduced-order models (ROMs) for the two-dimensional shallow water equation (SWE). The first one is based on the noncanonical Hamiltonian/Poisson form of the SWE. After integration in time by the fully implicit average vector field method, ROMs are constructed with proper orthogonal decomposition/discrete empirical interpolation me…
▽ More
In this paper, we present two different approaches for constructing reduced-order models (ROMs) for the two-dimensional shallow water equation (SWE). The first one is based on the noncanonical Hamiltonian/Poisson form of the SWE. After integration in time by the fully implicit average vector field method, ROMs are constructed with proper orthogonal decomposition/discrete empirical interpolation method (POD/DEIM) that preserves the Hamiltonian structure. In the second approach, the SWE as a partial differential equation with quadratic nonlinearity is integrated in time by the linearly implicit Kahan's method and ROMs are constructed with the tensorial POD that preserves the linear-quadratic structure of the SWE. We show that in both approaches, the invariants of the SWE such as the energy, enstrophy, mass, and circulation are preserved over a long period of time, leading to stable solutions. We conclude by demonstrating the accuracy and the computational efficiency of the reduced solutions by a numerical test problem.
△ Less
Submitted 22 July, 2020; v1 submitted 10 July, 2019;
originally announced July 2019.
-
Sum-of-squares optimization without semidefinite programming
Authors:
Dávid Papp,
Sercan Yıldız
Abstract:
We propose a homogeneous primal-dual interior-point method to solve sum-of-squares optimization problems by combining non-symmetric conic optimization techniques and polynomial interpolation. The approach optimizes directly over the sum-of-squares cone and its dual, circumventing the semidefinite programming (SDP) reformulation which requires a large number of auxiliary variables. As a result, it…
▽ More
We propose a homogeneous primal-dual interior-point method to solve sum-of-squares optimization problems by combining non-symmetric conic optimization techniques and polynomial interpolation. The approach optimizes directly over the sum-of-squares cone and its dual, circumventing the semidefinite programming (SDP) reformulation which requires a large number of auxiliary variables. As a result, it has substantially lower theoretical time and space complexity than the conventional SDP-based approach. Although our approach avoids the semidefinite programming reformulation, an optimal solution to the semidefinite program can be recovered with little additional effort. Computational results confirm that for problems involving high-degree polynomials, the proposed method is several orders of magnitude faster than semidefinite programming.
△ Less
Submitted 20 December, 2018; v1 submitted 5 December, 2017;
originally announced December 2017.
-
On "A Homogeneous Interior-Point Algorithm for Non-Symmetric Convex Conic Optimization"
Authors:
Dávid Papp,
Sercan Yıldız
Abstract:
In a recent paper, Skajaa and Ye proposed a homogeneous primal-dual interior-point method for non-symmetric conic optimization. The authors showed that their algorithm converges to $\varepsilon$-accuracy in $O(\sqrtν\log \varepsilon^{-1})$ iterations, where $ν$ is the complexity parameter associated with a barrier function for the primal cone, and thus achieves the best-known iteration complexity…
▽ More
In a recent paper, Skajaa and Ye proposed a homogeneous primal-dual interior-point method for non-symmetric conic optimization. The authors showed that their algorithm converges to $\varepsilon$-accuracy in $O(\sqrtν\log \varepsilon^{-1})$ iterations, where $ν$ is the complexity parameter associated with a barrier function for the primal cone, and thus achieves the best-known iteration complexity for this class of problems. However, an earlier result from the literature was used incorrectly in the proofs of two intermediate lemmas in that paper. In this note, we propose new proofs of these results, allowing the same complexity bound to be established.
△ Less
Submitted 14 June, 2018; v1 submitted 1 December, 2017;
originally announced December 2017.
-
New variations of power increasing sequences
Authors:
Sebnem Yildiz
Abstract:
The aim of this paper is to generalize a main theorem concerning weighted mean summability to absolute matrix summability which plays a vital role in summability theory and applications to the other sciences by using quasi-$f$-power sequences.
The aim of this paper is to generalize a main theorem concerning weighted mean summability to absolute matrix summability which plays a vital role in summability theory and applications to the other sciences by using quasi-$f$-power sequences.
△ Less
Submitted 13 November, 2017;
originally announced November 2017.
-
A matrix application of quasi-monotone sequences to Fourier series
Authors:
Sebnem Yildiz
Abstract:
In this paper, we have generalized a main theorem dealing with weighted mean summability method for absolute matrix summability method which plays a vital role in summability theory and applications to the other sciences by using quasi-monotone sequences. The main result in this paper extends the results in
In this paper, we have generalized a main theorem dealing with weighted mean summability method for absolute matrix summability method which plays a vital role in summability theory and applications to the other sciences by using quasi-monotone sequences. The main result in this paper extends the results in
△ Less
Submitted 17 October, 2017;
originally announced October 2017.
-
A new study on the absolute matrix summability of non-decreasing sequences
Authors:
Sebnem Yildiz
Abstract:
Recently, in [Bor4], Bor proved a main theorem dealing with $|\bar{N}, p_{n}|_{k}$ summability factors of infinite series. In the present paper, we have generalized that theorem for $|A, p_{n}|_{k}$ summability method by taking normal matrices in place of weighted mean matrices.
Recently, in [Bor4], Bor proved a main theorem dealing with $|\bar{N}, p_{n}|_{k}$ summability factors of infinite series. In the present paper, we have generalized that theorem for $|A, p_{n}|_{k}$ summability method by taking normal matrices in place of weighted mean matrices.
△ Less
Submitted 27 February, 2018; v1 submitted 29 September, 2017;
originally announced October 2017.
-
Two-Term Disjunctions on the Second-Order Cone
Authors:
Fatma Kilinc-Karzan,
Sercan Yildiz
Abstract:
Balas introduced disjunctive cuts in the 1970s for mixed-integer linear programs. Several recent papers have attempted to extend this work to mixed-integer conic programs. In this paper we study the structure of the convex hull of a two-term disjunction applied to the second-order cone, and develop a methodology to derive closed-form expressions for convex inequalities describing the resulting con…
▽ More
Balas introduced disjunctive cuts in the 1970s for mixed-integer linear programs. Several recent papers have attempted to extend this work to mixed-integer conic programs. In this paper we study the structure of the convex hull of a two-term disjunction applied to the second-order cone, and develop a methodology to derive closed-form expressions for convex inequalities describing the resulting convex hull. Our approach is based on first characterizing the structure of undominated valid linear inequalities for the disjunction and then using conic duality to derive a family of convex, possibly nonlinear, valid inequalities that correspond to these linear inequalities. We identify and study the cases where these valid inequalities can equivalently be expressed in conic quadratic form and where a single inequality from this family is sufficient to describe the convex hull. In particular, our results on two-term disjunctions on the second-order cone generalize related results on split cuts by Modaresi, Kilinc, and Vielma, and by Andersen and Jensen.
△ Less
Submitted 30 April, 2014;
originally announced April 2014.
