SRSvetlana Roudenko

Svetlana Roudenko

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Svetlana Roudenko has a total of 36 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Soliton Profiles in 1D: Enhanced Neural-Network Solvers and Constrained Deep Ritz Methods

    by Chandler Haight, Alex D. Rodriguez, Svetlana Roudenko

    We develop and compare {\it enhanced} neural-network solvers for one-dimensional ground-state profiles, which are solutions of $$Q^{\prime\prime}=bQ-\gamma Q^{p},\quad Q>0,\quad p>1, \quad b>0,$$the stationary profile equation for the nonlinear Schr\"odinger, generalized Korteweg--de Vries, and nonlinear Klein--Gordon equations. Its explicit $\sech$-type solutions serve as exact reference for error computation. We improve on the standard Adam-trained physics-informed neural network (PINN) baseline by introducing higher-order optimization via Gauss-Newton / Levenberg-Marquardt and L-BFGS refinements, as well as Fourier-feature PINNs, and constrained Deep Ritz (DR) formulations. In DR approaches we introduce two new neural-network formulations: a {\it Nehari-constrained} Deep Ritz method, in which the height is learned by projecting onto the Nehari/Pokhozhaev manifold, and a {\it Weinstein} Deep Ritz method, in which the profile shape minimizes the Weinstein quotient. The Deep Ritz formulations perform best overall, in particular, the Weinstein variant gives the best error-runtime balance. The neural network based solvers remain less accurate and efficient than classical Petviashvili iteration in 1D, but their mesh-free formulation and independence from exact profiles make them promising in higher dimensions and for more complicated nonlinearities and dispersive operators.

  • Local Well‐Posedness for the Biharmonic Nonlinear Schrödinger Equation in Low Dimensions

    by Iryna Petrenko, Oscar Riaño, Svetlana Roudenko

    In this paper we consider the fourth‐order (or biharmonic) nonlinear Schrödinger (NLS) equation in dimensions 1, 2, and 3, where the potential term is expressed as a power nonlinearity (for any positive power) and the dispersion operator has the fourth ‐order combined with the lower second ‐order. The fourth order NLS equation has recently attracted the attention of researchers since quartic solitons have been experimentally obtained in optics, and thus, the mathematical theory of solutions to the fourth‐order NLS equation in physical dimensions is timely to develop. In this work we show the local well‐possesses of the fourth‐order or biharmonic NLS equation on a weighted subset of a Sobolev space as well as in spaces.

  • The nonlinear Schrödinger equation with combined nonlinearities in 1D

    by Oscar Riaño, Alex D Rodriguez, Svetlana Roudenko

    We consider the one-dimensional nonlinear Schrödinger equation i u t + u x x + N ( u ) u = 0 , x , t ∈ R , with the nonlinearity term that is expressed as a sum of powers, possibly infinite: N ( u ) = ∑ d k | u | α k , α k > 0. We first investigate the local well-posedness of this equation for any positive powers of α k in a certain weighted class of initial data, subset of H 1 ( R ) . For that we use an approach of Cazenave–Naumkin (Cazenave and Naumkin 2016 Commun. Contemp. Math. 19 1650038), thus, avoiding any Strichartz estimates. Then, using the pseudo-conformal transformation, we extend the local result to the global one for the initial data with a quadratic phase. Furthermore, we investigate the asymptotic behaviour of such global solutions and prove scattering for data with the quadratic phase e i b | x | 2 with sufficiently large positive b , in H 1 ( R ) . One of the advantages of considering an infinite sum in the nonlinearity term is being able to consider exponential nonlinearities, such as e γ | u | k u , as well as sine or cosine nonlinearities, and obtain well-posedness in those cases, the first such result for most of those nonlinearities. To conclude, we show numerical simulations for various examples of combined nonlinearities, including the double nonlinearity and an exponential one, then investigate the behaviour of solutions with positive or negative initial b in a quadratic phase data. Furthermore, we also show that a ground state in the NLS equation with combined nonlinearities no longer provides a sharp threshold for global behaviour such as scattering vs finite time blow-up, instead the equation has a much richer dynamics.


  • Open Access

    SOLITON PROFILES IN 1D: CLASSICAL NUMERICAL SCHEMES VS. NEURAL NETWORK-BASED SOLVERS

    by Chandler Haight, Svetlana Roudenko, Zhongming Wang

    We present a comparative study of classical numerical solvers, such as Petviashvili’s method and finite difference with Newton iterations, and neural network-based methods for computing the solutions of a nonlinear elliptic equation, namely, soliton profiles to the dispersive PDEs that include the nonlinear Schro¨dinger, Klein-Gordon and the generalized KdV equations. To benchmark we consider the 1D setting and confirm that classical approaches retain high-order accuracy and strong computational efficiency for single-instance problems. Physics informed neural networks (PINNs) are also able to reproduce qualitative solutions but are generally less accurate and less efficient in low dimensions than classical solvers due to expensive training and slow convergence. We also investigate the operator-learning methods, which, although computationally intensive during training, can be reused across many parameter instances, providing rapid inference after pretraining, making them attractive for applications involving repeated simulations or real-time predictions. For single-instance computations the accuracy of operator-learning methods remains lower than that of classical methods or PINNs, in general.

  • Spectral property for the 2D Zakharov–Kuznetsov equation

    by S. Roudenko, J. Holmer

    We discuss a spectral property for the virial operator of the \(2\)D Zakharov–Kuznetsov (ZK) equation. This is a crucial ingredient to establish blow-up or asymptotic stability of solitary waves in higher-dimensional problems. This model in \(3\)D setting was originally introduced by Zakharov and Kuznetsov in plasma physics, and is also a higher-dimensional generalization of the well-known Korteweg–de Vries (KdV) equation. The problem of stability of solitary waves in ZK equation or stable blow-up in modified ZK (or KdV-type) equation is an important physical question, for which virial operators and their spectral properties are the essential elements of the analysis. In this paper we investigate this problem analytically and reduce it to verifying numerically only some signs of inner products and certain eigenvalues.


  • Open Access

    The nonlinear Schrödinger equation with combined nonlinearities in 1D

    by Oscar Riaño, Alex D Rodriguez, Svetlana Roudenko

    We consider the one-dimensional nonlinear Schrödinger equation $$ iu_t + u_{xx} + \mathcal{N}(u)u=0, \quad x,t \in \mathbb R, $$ with the nonlinearity term that is expressed as a sum of powers, possibly infinite: $$ \mathcal{N}(u) = \sum d_k |u|^{α_k}, \quad α_k > 0. $$ We first investigate the local well-posedness of this equation for any positive powers of $α_k$ in a certain weighted class of initial data, subset of $H^1 (\mathbb R)$. For that we use an approach of Cazenave-Naumkin [19], thus, avoiding any Strichartz estimates. Then, using the pseudo-conformal transformation, we extend the local result to the global one for the initial data with a quadratic phase. Furthermore, we investigate the asymptotic behavior of such global solutions and prove scattering for data with the quadratic phase $e^{ib|x|^2}$ with sufficiently large positive $b$, in $H^1(\mathbb R)$. One of the advantages of considering an infinite sum in the nonlinearity term is being able to consider exponential nonlinearities, such as $e^{γ|u|^{k}} u$, as well as sine or cosine nonlinearities, and obtain well-posedness in those cases, the first such result for most of those nonlinearities. To conclude, we show numerical simulations for various examples of combined nonlinearities, including the double nonlinearity and an exponential one, then investigate the behavior of solutions with positive or negative initial $b$ in a quadratic phase data. Furthermore, we also show that a ground state in the NLS equation with combined nonlinearities no longer provides a sharp threshold for global behavior such as scattering vs. finite time blow-up, instead the equation has a much richer dynamics.

  • Open Access

    Soliton profiles: Classical Numerical Schemes vs. Neural Network - Based Solvers

    by Chandler Haight, Svetlana Roudenko, Zhongming Wang

    We present a comparative study of classical numerical solvers, such as Petviashvili's method or finite difference with Newton iterations, and neural network-based methods for computing ground states or profiles of solitary-wave solutions to the one-dimensional dispersive PDEs that include the nonlinear Schrödinger, the nonlinear Klein-Gordon and the generalized KdV equations. We confirm that classical approaches retain high-order accuracy and strong computational efficiency for single-instance problems in the one-dimensional setting. Physics-informed neural networks (PINNs) are also able to reproduce qualitative solutions but are generally less accurate and less efficient in low dimensions than classical solvers due to expensive training and slow convergence. We also investigate the operator-learning methods, which, although computationally intensive during training, can be reused across many parameter instances, providing rapid inference after pretraining, making them attractive for applications involving repeated simulations or real-time predictions. For single-instance computations, however, the accuracy of operator-learning methods remains lower than that of classical methods or PINNs, in general.