-
Dynamics of ring solitons in an expanding cloud of a Bose-Einstein condensate
Authors:
A. M. Kamchatnov,
B. I. Suleimanov,
E. N. Tsoy
Abstract:
In this paper, we derive equations for the dynamics of ring dark solitons in an expanding cloud of a two-dimensional Bose-Einstein condensate. Assuming that the soliton's width is much smaller than its radius, we obtain the Hamilton equations for its evolution. Then they are transformed into the Newton equation, which is more convenient for applications. The general theory is illustrated by the so…
▽ More
In this paper, we derive equations for the dynamics of ring dark solitons in an expanding cloud of a two-dimensional Bose-Einstein condensate. Assuming that the soliton's width is much smaller than its radius, we obtain the Hamilton equations for its evolution. Then they are transformed into the Newton equation, which is more convenient for applications. The general theory is illustrated by the solution of the Newton equation for the case of the axially symmetric condensate cloud, which expands after switching off a harmonic trap. The validity of our approximate analytical approach is confirmed by comparison with the results of numerical simulations of the Gross-Pitaevskii equation.
△ Less
Submitted 2 April, 2025;
originally announced April 2025.
-
On the Impact of Performative Risk Minimization for Binary Random Variables
Authors:
Nikita Tsoy,
Ivan Kirev,
Negin Rahimiyazdi,
Nikola Konstantinov
Abstract:
Performativity, the phenomenon where outcomes are influenced by predictions, is particularly prevalent in social contexts where individuals strategically respond to a deployed model. In order to preserve the high accuracy of machine learning models under distribution shifts caused by performativity, Perdomo et al. (2020) introduced the concept of performative risk minimization (PRM). While this fr…
▽ More
Performativity, the phenomenon where outcomes are influenced by predictions, is particularly prevalent in social contexts where individuals strategically respond to a deployed model. In order to preserve the high accuracy of machine learning models under distribution shifts caused by performativity, Perdomo et al. (2020) introduced the concept of performative risk minimization (PRM). While this framework ensures model accuracy, it overlooks the impact of the PRM on the underlying distributions and the predictions of the model. In this paper, we initiate the analysis of the impact of PRM, by studying performativity for a sequential performative risk minimization problem with binary random variables and linear performative shifts. We formulate two natural measures of impact. In the case of full information, where the distribution dynamics are known, we derive explicit formulas for the PRM solution and our impact measures. In the case of partial information, we provide performative-aware statistical estimators, as well as simulations. Our analysis contrasts PRM to alternatives that do not model data shift and indicates that PRM can have amplified side effects compared to such methods.
△ Less
Submitted 4 February, 2025;
originally announced February 2025.
-
Incentivizing Truthful Collaboration in Heterogeneous Federated Learning
Authors:
Dimitar Chakarov,
Nikita Tsoy,
Kristian Minchev,
Nikola Konstantinov
Abstract:
Federated learning (FL) is a distributed collaborative learning method, where multiple clients learn together by sharing gradient updates instead of raw data. However, it is well-known that FL is vulnerable to manipulated updates from clients. In this work we study the impact of data heterogeneity on clients' incentives to manipulate their updates. First, we present heterogeneous collaborative lea…
▽ More
Federated learning (FL) is a distributed collaborative learning method, where multiple clients learn together by sharing gradient updates instead of raw data. However, it is well-known that FL is vulnerable to manipulated updates from clients. In this work we study the impact of data heterogeneity on clients' incentives to manipulate their updates. First, we present heterogeneous collaborative learning scenarios where a client can modify their updates to be better off, and show that these manipulations can lead to diminishing model performance. To prevent such modifications, we formulate a game in which clients may misreport their gradient updates in order to "steer" the server model to their advantage. We develop a payment rule that provably disincentivizes sending modified updates under the FedSGD protocol. We derive explicit bounds on the clients' payments and the convergence rate of the global model, which allows us to study the trade-off between heterogeneity, payments and convergence. Finally, we provide an experimental evaluation of the effectiveness of our payment rule in the FedSGD, median-based aggregation FedSGD and FedAvg protocols on three tasks in computer vision and natural language processing. In all cases we find that our scheme successfully disincentivizes modifications.
△ Less
Submitted 5 March, 2025; v1 submitted 1 December, 2024;
originally announced December 2024.
-
Generic quartic solitons in optical media
Authors:
Eduard N. Tsoy,
Laziz A. Suyunov
Abstract:
Our analysis suggests strongly that stationary pulses exist in nonlinear media with second-, third-, and fourth-order dispersion. A theory, based on the variational approach, is developed for finding approximate parameters of such solitons. It is obtained that the soliton velocity in the retarded reference frame can be different from the inverse of the group velocity of linear waves. It is shown t…
▽ More
Our analysis suggests strongly that stationary pulses exist in nonlinear media with second-, third-, and fourth-order dispersion. A theory, based on the variational approach, is developed for finding approximate parameters of such solitons. It is obtained that the soliton velocity in the retarded reference frame can be different from the inverse of the group velocity of linear waves. It is shown that the interaction of the pulse spectrum with that of linear waves can affect the existence of stationary solitons. These theoretical results are supported by numerical simulations. Transformations between solitons of different systems are derived. A generalization for solitons in media with the highest even-order dispersion is suggested.
△ Less
Submitted 28 May, 2024;
originally announced May 2024.
-
Simplicity Bias of Two-Layer Networks beyond Linearly Separable Data
Authors:
Nikita Tsoy,
Nikola Konstantinov
Abstract:
Simplicity bias, the propensity of deep models to over-rely on simple features, has been identified as a potential reason for limited out-of-distribution generalization of neural networks (Shah et al., 2020). Despite the important implications, this phenomenon has been theoretically confirmed and characterized only under strong dataset assumptions, such as linear separability (Lyu et al., 2021). I…
▽ More
Simplicity bias, the propensity of deep models to over-rely on simple features, has been identified as a potential reason for limited out-of-distribution generalization of neural networks (Shah et al., 2020). Despite the important implications, this phenomenon has been theoretically confirmed and characterized only under strong dataset assumptions, such as linear separability (Lyu et al., 2021). In this work, we characterize simplicity bias for general datasets in the context of two-layer neural networks initialized with small weights and trained with gradient flow. Specifically, we prove that in the early training phases, network features cluster around a few directions that do not depend on the size of the hidden layer. Furthermore, for datasets with an XOR-like pattern, we precisely identify the learned features and demonstrate that simplicity bias intensifies during later training stages. These results indicate that features learned in the middle stages of training may be more useful for OOD transfer. We support this hypothesis with experiments on image data.
△ Less
Submitted 7 November, 2024; v1 submitted 27 May, 2024;
originally announced May 2024.
-
Provable Mutual Benefits from Federated Learning in Privacy-Sensitive Domains
Authors:
Nikita Tsoy,
Anna Mihalkova,
Teodora Todorova,
Nikola Konstantinov
Abstract:
Cross-silo federated learning (FL) allows data owners to train accurate machine learning models by benefiting from each others private datasets. Unfortunately, the model accuracy benefits of collaboration are often undermined by privacy defenses. Therefore, to incentivize client participation in privacy-sensitive domains, a FL protocol should strike a delicate balance between privacy guarantees an…
▽ More
Cross-silo federated learning (FL) allows data owners to train accurate machine learning models by benefiting from each others private datasets. Unfortunately, the model accuracy benefits of collaboration are often undermined by privacy defenses. Therefore, to incentivize client participation in privacy-sensitive domains, a FL protocol should strike a delicate balance between privacy guarantees and end-model accuracy. In this paper, we study the question of when and how a server could design a FL protocol provably beneficial for all participants. First, we provide necessary and sufficient conditions for the existence of mutually beneficial protocols in the context of mean estimation and convex stochastic optimization. We also derive protocols that maximize the total clients' utility, given symmetric privacy preferences. Finally, we design protocols maximizing end-model accuracy and demonstrate their benefits in synthetic experiments.
△ Less
Submitted 7 November, 2024; v1 submitted 11 March, 2024;
originally announced March 2024.
-
Strategic Data Sharing between Competitors
Authors:
Nikita Tsoy,
Nikola Konstantinov
Abstract:
Collaborative learning techniques have significantly advanced in recent years, enabling private model training across multiple organizations. Despite this opportunity, firms face a dilemma when considering data sharing with competitors -- while collaboration can improve a company's machine learning model, it may also benefit competitors and hence reduce profits. In this work, we introduce a genera…
▽ More
Collaborative learning techniques have significantly advanced in recent years, enabling private model training across multiple organizations. Despite this opportunity, firms face a dilemma when considering data sharing with competitors -- while collaboration can improve a company's machine learning model, it may also benefit competitors and hence reduce profits. In this work, we introduce a general framework for analyzing this data-sharing trade-off. The framework consists of three components, representing the firms' production decisions, the effect of additional data on model quality, and the data-sharing negotiation process, respectively. We then study an instantiation of the framework, based on a conventional market model from economic theory, to identify key factors that affect collaboration incentives. Our findings indicate a profound impact of market conditions on the data-sharing incentives. In particular, we find that reduced competition, in terms of the similarities between the firms' products, and harder learning tasks foster collaboration.
△ Less
Submitted 30 October, 2023; v1 submitted 25 May, 2023;
originally announced May 2023.
-
Modulational instability and quantum droplets in a two-dimensional Bose-Einstein condensate
Authors:
Sherzod R. Otajonov,
Eduard N. Tsoy,
Fatkhulla Kh. Abdullaev
Abstract:
Modulational instability of a uniform two-dimensional binary Bose-Einstein condensate (BEC) in the presence of quantum fluctuations is studied. The analysis is based on the coupled Gross-Pitaevskii equations. It is shown that quantum fluctuations can induce instability when the BEC density is below a threshold. The dependence of the growth rate of modulations on the BEC parameters is found. It is…
▽ More
Modulational instability of a uniform two-dimensional binary Bose-Einstein condensate (BEC) in the presence of quantum fluctuations is studied. The analysis is based on the coupled Gross-Pitaevskii equations. It is shown that quantum fluctuations can induce instability when the BEC density is below a threshold. The dependence of the growth rate of modulations on the BEC parameters is found. It is observed that an asymmetry of the interaction parameters and/or initial densities of the components typically decreases the growth rate. Further development of the instability results in a break-up of the BEC into a set of quantum droplets. These droplets merge dynamically with each other so that the total number of droplets decreases rapidly. The rate of this decrease is evaluated numerically for different initial parameters.
△ Less
Submitted 1 July, 2022;
originally announced July 2022.
-
The variational approximation for two-dimensional quantum droplets
Authors:
Sherzod R. Otajonov,
Eduard N. Tsoy,
Fatkhulla Kh. Abdullaev
Abstract:
The dynamics of a two-dimensional Bose-Einstein condensate in a presence of quantum fluctuations is studied. The properties of localized density distributions, quantum droplets (QDs), are analyzed by means of the variational approach. It is demonstrated that the super-Gaussian function gives a good approximation for profiles of fundamental QDs and droplets with non-zero vorticity. The dynamical eq…
▽ More
The dynamics of a two-dimensional Bose-Einstein condensate in a presence of quantum fluctuations is studied. The properties of localized density distributions, quantum droplets (QDs), are analyzed by means of the variational approach. It is demonstrated that the super-Gaussian function gives a good approximation for profiles of fundamental QDs and droplets with non-zero vorticity. The dynamical equations for parameters of QDs are obtained. Fixed points of these equations determine the parameters of stationary QDs. The period of small oscillations of QDs near the stationary state is estimated. It is obtained that periodic modulations of the strength of quantum fluctuations can actuate different processes, including resonance oscillations of the QD parameters, an emission of waves and a splitting of QDs into smaller droplets.
△ Less
Submitted 24 December, 2020;
originally announced December 2020.
-
Stationary and dynamical properties of one-dimensional quantum droplets
Authors:
Sherzod R. Otajonov,
Eduard N. Tsoy,
Fatkhulla Kh. Abdullaev
Abstract:
The dynamics of quantum droplets in 1D is analyzed on the basis of the variational approach (VA). It is shown that the VA based on the super-Gaussian function gives a good approximation of stationary states. The period of small oscillations of the perturbed droplet is obtained. It is found numerically that oscillations are almost undamped for many periods. Based on the VA, an existence of stable l…
▽ More
The dynamics of quantum droplets in 1D is analyzed on the basis of the variational approach (VA). It is shown that the VA based on the super-Gaussian function gives a good approximation of stationary states. The period of small oscillations of the perturbed droplet is obtained. It is found numerically that oscillations are almost undamped for many periods. Based on the VA, an existence of stable localized states for different combinations of signs of nonlinearities is demonstrated.
△ Less
Submitted 21 September, 2019;
originally announced September 2019.
-
Introduction to nonlinear discrete systems: Theory and modeling
Authors:
E. N. Tsoy,
B. A. Umarov
Abstract:
An analysis of discrete systems is important for understanding of various physical processes, such as excitations in crystal lattices and molecular chains, the light propagation in waveguide arrays, and the dynamics of Bose-condensate droplets. In basic physical courses, usually linear properties of discrete systems are studied. In this paper we propose a pedagogical introduction to the theory of…
▽ More
An analysis of discrete systems is important for understanding of various physical processes, such as excitations in crystal lattices and molecular chains, the light propagation in waveguide arrays, and the dynamics of Bose-condensate droplets. In basic physical courses, usually linear properties of discrete systems are studied. In this paper we propose a pedagogical introduction to the theory of nonlinear distributed systems. The main ideas and methods are illustrated using a universal model for different physical applications, the discrete nonlinear Schrödinger (DNLS) equation. We consider solutions of the DNLS equation and analyze their linear stability. The notions of nonlinear plane waves, modulational instability, discrete solitons and the anti-continuum limit are introduced and thoroughly discussed. A Mathematica program is provided for better comprehension of results and further exploration. Also, few problems, extending the topic of the paper, for independent solution are given.
△ Less
Submitted 5 August, 2019;
originally announced August 2019.
-
Dynamics of fronts in optical media with linear gain and nonlinear losses
Authors:
Izzat M. Allayarov,
Eduard N. Tsoy
Abstract:
The dynamics of fronts, or kinks, in dispersive media with gain and losses is considered. It is shown that the front parameters, such as the velocity and width, depend on initial conditions. This result is not typical for dissipative systems. For exponentially decreasing initial conditions, the relations for the front parameters are found. A presence of the global bifurcation, when a soliton solut…
▽ More
The dynamics of fronts, or kinks, in dispersive media with gain and losses is considered. It is shown that the front parameters, such as the velocity and width, depend on initial conditions. This result is not typical for dissipative systems. For exponentially decreasing initial conditions, the relations for the front parameters are found. A presence of the global bifurcation, when a soliton solution is replaced by the front solution, is demonstrated. It is also shown that in order to observe fronts, the front velocity should be larger than the characteristic velocity of the modulational instability.
△ Less
Submitted 5 August, 2014;
originally announced August 2014.
-
Stable localized modes in asymmetric waveguides with gain and loss
Authors:
Eduard N. Tsoy,
Izzat M. Allayarov,
Fatkhulla Kh. Abdullaev
Abstract:
It is shown that asymmetric waveguides with gain and loss can support a stable propagation of optical beams. This means that the propagation constants of modes of the corresponding complex optical potential are real. A class of such waveguides is found from a relation between two spectral problems. A particular example of an asymmetric waveguide, described by the hyperbolic functions, is analyzed.…
▽ More
It is shown that asymmetric waveguides with gain and loss can support a stable propagation of optical beams. This means that the propagation constants of modes of the corresponding complex optical potential are real. A class of such waveguides is found from a relation between two spectral problems. A particular example of an asymmetric waveguide, described by the hyperbolic functions, is analyzed. The existence and stability of linear modes and of continuous families of nonlinear modes are demonstrated.
△ Less
Submitted 5 August, 2014;
originally announced August 2014.
-
Vortex flow in the technology of radiation wave cracking (RWC)
Authors:
L. A. Tsoy,
V. N. Kolushov,
A. G. Komarov,
A. N. Tsoy
Abstract:
This article examines the theory of vortex flows in relation to the processes occurring in the radiation-wave cracking of crude oil, when the crude oil is sprayed into the gas stream in the form of a mist and then is fed into the reactor, where it is treated by the accelerated electrons and the UHF radiation. The output of this process are the products with the specified parameters (high-octane pe…
▽ More
This article examines the theory of vortex flows in relation to the processes occurring in the radiation-wave cracking of crude oil, when the crude oil is sprayed into the gas stream in the form of a mist and then is fed into the reactor, where it is treated by the accelerated electrons and the UHF radiation. The output of this process are the products with the specified parameters (high-octane petroleum products). This process operates at the ambient pressure and temperature, which makes the process safer for industrial purposes. Besides the process itself, the authors described the equipment used in this process, as well as the parameters of the optimal process.
△ Less
Submitted 16 September, 2012;
originally announced September 2012.
-
Bifurcations from stationary to pulsating solitons in the cubic-quintic complex Ginzburg-Landau equation
Authors:
E. N. Tsoy,
N. Akhmediev
Abstract:
Stationary to pulsating soliton bifurcation analysis of the complex Ginzburg-Landau equation (CGLE) is presented. The analysis is based on a reduction from an infinite-dimensional dynamical dissipative system to a finite-dimensional model. Stationary solitons, with constant amplitude and width, are associated with fixed points in the model. For the first time, pulsating solitons are shown to be…
▽ More
Stationary to pulsating soliton bifurcation analysis of the complex Ginzburg-Landau equation (CGLE) is presented. The analysis is based on a reduction from an infinite-dimensional dynamical dissipative system to a finite-dimensional model. Stationary solitons, with constant amplitude and width, are associated with fixed points in the model. For the first time, pulsating solitons are shown to be stable limit cycles in the finite-dimensional dynamical system. The boundaries between the two types of solutions are obtained approximately from the reduced model. These boundaries are reasonably close to those predicted by direct numerical simulations of the CGLE.
△ Less
Submitted 14 February, 2006;
originally announced February 2006.
-
Array of Bose-Einstein condensates under time-periodic Feshbach-resonance management
Authors:
F. Kh. Abdullaev,
E. N. Tsoy,
B. A. Malomed,
R. A. Kraenkel
Abstract:
The dynamics of a discrete soliton in an array of Bose-Einstein condensates under the action of a periodically time-modulated atomic scattering length (``Feshbach-resonance management, FRM'') is investigated. The cases of both slow and rapid modulation, in comparison with the tunneling frequency, are considered. We employ a discrete variational approach for the analysis of the system. The existe…
▽ More
The dynamics of a discrete soliton in an array of Bose-Einstein condensates under the action of a periodically time-modulated atomic scattering length (``Feshbach-resonance management, FRM'') is investigated. The cases of both slow and rapid modulation, in comparison with the tunneling frequency, are considered. We employ a discrete variational approach for the analysis of the system. The existence of nonlinear resonances and chaos is predicted at special values of the driving frequency. Soliton splitting is observed in numerical simulations. In the case of the rapid modulation, we derive an averaged equation, which is a generalized discrete nonlinear Schroedinger equation, including higher-order effective nonlinearities and intersite nonlinear interactions. Thus the predicted discrete FRM solitons are a direct matter-wave analog of recently investigated discrete diffraction-managed optical solitons.
△ Less
Submitted 11 June, 2003;
originally announced June 2003.
-
Interaction of pulses in nonlinear Schroedinger model
Authors:
E. N. Tsoy,
F. Kh. Abdullaev
Abstract:
The interaction of two rectangular pulses in nonlinear Schroedinger model is studied by solving the appropriate Zakharov-Shabat system. It is shown that two real pulses may result in appearance of moving solitons. Different limiting cases, such as a single pulse with a phase jump, a single chirped pulse, in-phase and out-of-phase pulses, and pulses with frequency separation, are analyzed. The th…
▽ More
The interaction of two rectangular pulses in nonlinear Schroedinger model is studied by solving the appropriate Zakharov-Shabat system. It is shown that two real pulses may result in appearance of moving solitons. Different limiting cases, such as a single pulse with a phase jump, a single chirped pulse, in-phase and out-of-phase pulses, and pulses with frequency separation, are analyzed. The thresholds of creation of new solitons and multi-soliton states are found.
△ Less
Submitted 26 March, 2003;
originally announced March 2003.
