-
Data-driven Identification of Attractors Using Machine Learning
Authors:
Marcio Gameiro,
Brittany Gelb,
William Kalies,
Miroslav Kramar,
Konstantin Mischaikow,
Paul Tatasciore
Abstract:
In this paper we explore challenges in developing a topological framework in which machine learning can be used to robustly characterize global dynamics. Specifically, we focus on learning a useful discretization of the phase space of a flow on compact, hyperrectangle in $\mathbb{R}^n$ from a neural network trained on labeled orbit data. A characterization of the structure of the global dynamics i…
▽ More
In this paper we explore challenges in developing a topological framework in which machine learning can be used to robustly characterize global dynamics. Specifically, we focus on learning a useful discretization of the phase space of a flow on compact, hyperrectangle in $\mathbb{R}^n$ from a neural network trained on labeled orbit data. A characterization of the structure of the global dynamics is obtained from approximations of attracting neighborhoods provided by the phase space discretization. The perspective that motivates this work is based on Conley's topological approach to dynamics, which provides a means to evaluate the efficacy and efficiency of our approach.
△ Less
Submitted 6 June, 2025;
originally announced June 2025.
-
Rigorously Characterizing Dynamics with Machine Learning
Authors:
Marcio Gameiro,
Brittany Gelb,
Konstantin Mischaikow
Abstract:
The identification of dynamics from time series data is a problem of general interest. It is well established that dynamics on the level of invariant sets, the primary objects of interest in the classical theory of dynamical systems, is not computable. We recall a coarser characterization of dynamics based on order theory and algebraic topology and prove that this characterization can be identifie…
▽ More
The identification of dynamics from time series data is a problem of general interest. It is well established that dynamics on the level of invariant sets, the primary objects of interest in the classical theory of dynamical systems, is not computable. We recall a coarser characterization of dynamics based on order theory and algebraic topology and prove that this characterization can be identified using approximations that can be realized by feedforward neural networks.
△ Less
Submitted 22 May, 2025;
originally announced May 2025.
-
Global Dynamics of Ordinary Differential Equations: Wall Labelings, Conley Complexes, and Ramp Systems
Authors:
Marcio Gameiro,
Tomáš Gedeon,
Hiroshi Kokubu,
Konstantin Mischaikow,
Hiroe Oka,
Bernardo Rivas,
Ewerton Vieira,
Daniel Gameiro
Abstract:
We introduce a combinatorial topological framework for characterizing the global dynamics of ordinary differential equations (ODEs). The approach is motivated by the study of gene regulatory networks, which are often modeled by ODEs that are not explicitly derived from first principles.
The proposed method involves constructing a combinatorial model from a set of parameters and then embedding th…
▽ More
We introduce a combinatorial topological framework for characterizing the global dynamics of ordinary differential equations (ODEs). The approach is motivated by the study of gene regulatory networks, which are often modeled by ODEs that are not explicitly derived from first principles.
The proposed method involves constructing a combinatorial model from a set of parameters and then embedding the model into a continuous setting in such a way that the algebraic topological invariants are preserved. In this manuscript, we build upon the software Dynamic Signatures Generated by Regulatory Networks (DSGRN), a software package that is used to explore the dynamics generated by a regulatory network. By extending its functionalities, we deduce the global dynamical information of the ODE and extract information regarding equilibria, periodic orbits, connecting orbits and bifurcations.
We validate our results through algebraic topological tools and analytical bounds, and the effectiveness of this framework is demonstrated through several examples and possible future directions.
△ Less
Submitted 15 December, 2024;
originally announced December 2024.
-
Inferring Long-term Dynamics of Ecological Communities Using Combinatorics
Authors:
William S. Cuello,
Marcio Gameiro,
Juan A. Bonachela,
Konstantin Mischaikow
Abstract:
In an increasingly changing world, predicting the fate of species across the globe has become a major concern. Understanding how the population dynamics of various species and communities will unfold requires predictive tools that experimental data alone can not capture. Here, we introduce our combinatorial framework, Widespread Ecological Networks and their Dynamical Signatures (WENDyS) which, us…
▽ More
In an increasingly changing world, predicting the fate of species across the globe has become a major concern. Understanding how the population dynamics of various species and communities will unfold requires predictive tools that experimental data alone can not capture. Here, we introduce our combinatorial framework, Widespread Ecological Networks and their Dynamical Signatures (WENDyS) which, using data on the relative strengths of interactions and growth rates within a community of species predicts all possible long-term outcomes of the community. To this end, WENDyS partitions the multidimensional parameter space (formed by the strengths of interactions and growth rates) into a finite number of regions, each corresponding to a unique set of coarse population dynamics. Thus, WENDyS ultimately creates a library of all possible outcomes for the community. On the one hand, our framework avoids the typical ``parameter sweeps'' that have become ubiquitous across other forms of mathematical modeling, which can be computationally expensive for ecologically realistic models and examples. On the other hand, WENDyS opens the opportunity for interdisciplinary teams to use standard experimental data (i.e., strengths of interactions and growth rates) to filter down the possible end states of a community. To demonstrate the latter, here we present a case study from the Indonesian Coral Reef. We analyze how different interactions between anemone and anemonefish species lead to alternative stable states for the coral reef community, and how competition can increase the chance of exclusion for one or more species. WENDyS, thus, can be used to anticipate ecological outcomes and test the effectiveness of management (e.g., conservation) strategies.
△ Less
Submitted 1 September, 2023;
originally announced September 2023.
-
A framework for rigorous computational methods using Haar wavelets for differential equations
Authors:
Guilherme Nakassima,
Marcio Gameiro
Abstract:
This work presents a framework for a-posteriori error-estimating algorithms for differential equations which combines the radii polynomial approach with Haar wavelets. By using Haar wavelets, we obtain recursive structures for the matrix representations of the differential operators and quadratic nonlinearities, which can be exploited for the radii polynomial method in order to get error estimates…
▽ More
This work presents a framework for a-posteriori error-estimating algorithms for differential equations which combines the radii polynomial approach with Haar wavelets. By using Haar wavelets, we obtain recursive structures for the matrix representations of the differential operators and quadratic nonlinearities, which can be exploited for the radii polynomial method in order to get error estimates in the $L^2$ sense. This allows the method to be applicable when the system or solution is not continuous, which is a limitation of other radii-polynomial-based methods. Numerical examples show how the method is implemented in practice.
△ Less
Submitted 27 April, 2023;
originally announced April 2023.
-
A yeast cell cycle pulse generator model shows consistency with multiple oscillatory and checkpoint mutant datasets
Authors:
Julian Fox,
Breschine Cummins,
Robert C. Moseley,
Marcio Gameiro,
Steven B. Haase
Abstract:
The regulatory mechanisms driving progression of the yeast cell cycle appears to be comprised of an interacting network of transcription factors (TFs), cyclin-dependent kinases (CDK) and ubiquitin ligases. From a systems perspective the controlling regulatory network must produce robust periodic behavior during proliferative phases, but have the capability to halt the cycle when unfavorable condit…
▽ More
The regulatory mechanisms driving progression of the yeast cell cycle appears to be comprised of an interacting network of transcription factors (TFs), cyclin-dependent kinases (CDK) and ubiquitin ligases. From a systems perspective the controlling regulatory network must produce robust periodic behavior during proliferative phases, but have the capability to halt the cycle when unfavorable conditions trigger a checkpoint arrest. How the individual components of the network contribute to these dynamical phenotypes remains an open question. Here we evaluate the capability of a simplified network model hypothesized to contain key elements of the regulation of cell-cycle progression to reproduce observed transcriptomic behaviors. We match time-series data from both cycling and checkpoint arrested cells to the predictions of a relatively simple cell-cycle network model using an asynchronous multi-level Boolean approach. We show that this single network model, despite its simplicity, is capable of exhibiting dynamical behavior similar to the datasets in most cases, and where it does not, we identified hypotheses that suggest missing components of the network.
△ Less
Submitted 24 February, 2023;
originally announced February 2023.
-
Identifying Nonlinear Dynamics with High Confidence from Sparse Data
Authors:
Bogdan Batko,
Marcio Gameiro,
Ying Hung,
William Kalies,
Konstantin Mischaikow,
Ewerton Vieira
Abstract:
We introduce a novel procedure that, given sparse data generated from a stationary deterministic nonlinear dynamical system, can characterize specific local and/or global dynamic behavior with rigorous probability guarantees. More precisely, the sparse data is used to construct a statistical surrogate model based on a Gaussian process (GP). The dynamics of the surrogate model is interrogated using…
▽ More
We introduce a novel procedure that, given sparse data generated from a stationary deterministic nonlinear dynamical system, can characterize specific local and/or global dynamic behavior with rigorous probability guarantees. More precisely, the sparse data is used to construct a statistical surrogate model based on a Gaussian process (GP). The dynamics of the surrogate model is interrogated using combinatorial methods and characterized using algebraic topological invariants (Conley index). The GP predictive distribution provides a lower bound on the confidence that these topological invariants, and hence the characterized dynamics, apply to the unknown dynamical system (a sample path of the GP). The focus of this paper is on explaining the ideas, thus we restrict our examples to one-dimensional systems and show how to capture the existence of fixed points, periodic orbits, connecting orbits, bistability, and chaotic dynamics.
△ Less
Submitted 15 September, 2023; v1 submitted 28 June, 2022;
originally announced June 2022.
-
Towards computational Morse-Floer homology: forcing results for connecting orbits by computing relative indices of critical points
Authors:
Jan Bouwe van den Berg,
Marcio Gameiro,
Jean-Philippe Lessard,
Rob van der Vorst
Abstract:
To make progress towards better computability of Morse-Floer homology, and thus enhance the applicability of Floer theory, it is essential to have tools to determine the relative index of equilibria. Since even the existence of nontrivial stationary points is often difficult to accomplish, extracting their index information is usually out of reach. In this paper we establish a computer-assisted pr…
▽ More
To make progress towards better computability of Morse-Floer homology, and thus enhance the applicability of Floer theory, it is essential to have tools to determine the relative index of equilibria. Since even the existence of nontrivial stationary points is often difficult to accomplish, extracting their index information is usually out of reach. In this paper we establish a computer-assisted proof approach to determining relative indices of stationary states. We introduce the general framework and then focus on three example problems described by partial differential equations to show how these ideas work in practice. Based on a rigorous implementation, with accompanying code made available, we determine the relative indices of many stationary points. Moreover, we show how forcing results can be then used to prove theorems about connecting orbits and traveling waves in partial differential equations.
△ Less
Submitted 27 October, 2023; v1 submitted 18 June, 2022;
originally announced June 2022.
-
Morse Graphs: Topological Tools for Analyzing the Global Dynamics of Robot Controllers
Authors:
Ewerton R. Vieira,
Edgar Granados,
Aravind Sivaramakrishnan,
Marcio Gameiro,
Konstantin Mischaikow,
Kostas E. Bekris
Abstract:
Understanding the global dynamics of a robot controller, such as identifying attractors and their regions of attraction (RoA), is important for safe deployment and synthesizing more effective hybrid controllers. This paper proposes a topological framework to analyze the global dynamics of robot controllers, even data-driven ones, in an effective and explainable way. It builds a combinatorial repre…
▽ More
Understanding the global dynamics of a robot controller, such as identifying attractors and their regions of attraction (RoA), is important for safe deployment and synthesizing more effective hybrid controllers. This paper proposes a topological framework to analyze the global dynamics of robot controllers, even data-driven ones, in an effective and explainable way. It builds a combinatorial representation representing the underlying system's state space and non-linear dynamics, which is summarized in a directed acyclic graph, the Morse graph. The approach only probes the dynamics locally by forward propagating short trajectories over a state-space discretization, which needs to be a Lipschitz-continuous function. The framework is evaluated given either numerical or data-driven controllers for classical robotic benchmarks. It is compared against established analytical and recent machine learning alternatives for estimating the RoAs of such controllers. It is shown to outperform them in accuracy and efficiency. It also provides deeper insights as it describes the global dynamics up to the discretization's resolution. This allows to use the Morse graph to identify how to synthesize controllers to form improved hybrid solutions or how to identify the physical limitations of a robotic system.
△ Less
Submitted 13 August, 2022; v1 submitted 16 February, 2022;
originally announced February 2022.
-
Extending combinatorial regulatory network modeling to include activity control and decay modulation
Authors:
Bree Cummins,
Marcio Gameiro,
Tomas Gedeon,
Shane Kepley,
Konstantin Mischaikow,
Lun Zhang
Abstract:
Understanding how the structure of within-system interactions affects the dynamics of the system is important in many areas of science. We extend a network dynamics modeling platform DSGRN, which combinatorializes both dynamics and parameter space to construct finite but accurate summaries of network dynamics, to new types of interactions. While the standard DSGRN assumes that each network edge co…
▽ More
Understanding how the structure of within-system interactions affects the dynamics of the system is important in many areas of science. We extend a network dynamics modeling platform DSGRN, which combinatorializes both dynamics and parameter space to construct finite but accurate summaries of network dynamics, to new types of interactions. While the standard DSGRN assumes that each network edge controls the rate of abundance of the target node, the new edges may control either activity level or a decay rate of its target. While motivated by processes of post-transcriptional modification and ubiquitination in systems biology, our extension is applicable to the dynamics of any signed directed network.
△ Less
Submitted 16 May, 2022; v1 submitted 2 November, 2021;
originally announced November 2021.
-
Extracting Global Dynamics of Loss Landscape in Deep Learning Models
Authors:
Mohammed Eslami,
Hamed Eramian,
Marcio Gameiro,
William Kalies,
Konstantin Mischaikow
Abstract:
Deep learning models evolve through training to learn the manifold in which the data exists to satisfy an objective. It is well known that evolution leads to different final states which produce inconsistent predictions of the same test data points. This calls for techniques to be able to empirically quantify the difference in the trajectories and highlight problematic regions. While much focus is…
▽ More
Deep learning models evolve through training to learn the manifold in which the data exists to satisfy an objective. It is well known that evolution leads to different final states which produce inconsistent predictions of the same test data points. This calls for techniques to be able to empirically quantify the difference in the trajectories and highlight problematic regions. While much focus is placed on discovering what models learn, the question of how a model learns is less studied beyond theoretical landscape characterizations and local geometric approximations near optimal conditions. Here, we present a toolkit for the Dynamical Organization Of Deep Learning Loss Landscapes, or DOODL3. DOODL3 formulates the training of neural networks as a dynamical system, analyzes the learning process, and presents an interpretable global view of trajectories in the loss landscape. Our approach uses the coarseness of topology to capture the granularity of geometry to mitigate against states of instability or elongated training. Overall, our analysis presents an empirical framework to extract the global dynamics of a model and to use that information to guide the training of neural networks.
△ Less
Submitted 14 June, 2021;
originally announced June 2021.
-
Rational design of complex phenotype via network models
Authors:
Marcio Gameiro,
Tomas Gedeon,
Shane Kepley,
Konstantin Mischaikow
Abstract:
We demonstrate a modeling and computational framework that allows for rapid screening of thousands of potential network designs for particular dynamic behavior. To illustrate this capability we consider the problem of hysteresis, a prerequisite for construction of robust bistable switches and hence a cornerstone for construction of more complex synthetic circuits. We evaluate and rank most three n…
▽ More
We demonstrate a modeling and computational framework that allows for rapid screening of thousands of potential network designs for particular dynamic behavior. To illustrate this capability we consider the problem of hysteresis, a prerequisite for construction of robust bistable switches and hence a cornerstone for construction of more complex synthetic circuits. We evaluate and rank most three node networks according to their ability to robustly exhibit hysteresis where robustness is measured with respect to parameters over multiple dynamic phenotypes. Focusing on the highest ranked networks, we demonstrate how additional robustness and design constraints can be applied. We compare our results to more traditional methods based on specific parameterization of ordinary differential equation models and demonstrate a strong qualitative match at a small fraction of the computational cost.
△ Less
Submitted 21 May, 2021; v1 submitted 8 October, 2020;
originally announced October 2020.
-
On the feasibility and convergence of the inexact Newton method under minor conditions on the error terms
Authors:
Eduardo Ramos,
Marcio Gameiro,
Victor Nolasco
Abstract:
In this paper we introduce a semi-local theorem for the feasibility and convergence of the inexact Newton method, regarding the sequence $x_{k+1} = x_k - Df(x_k)^{-1}f(x_k) + r_k$, where $r_k$ represents the error in each step. Unlike the previous results of this type in the literature, we prove the feasibility of the inexact Newton method under the minor hypothesis that the error $r_k$ is bounded…
▽ More
In this paper we introduce a semi-local theorem for the feasibility and convergence of the inexact Newton method, regarding the sequence $x_{k+1} = x_k - Df(x_k)^{-1}f(x_k) + r_k$, where $r_k$ represents the error in each step. Unlike the previous results of this type in the literature, we prove the feasibility of the inexact Newton method under the minor hypothesis that the error $r_k$ is bounded by a small constant to be computed, and moreover we prove results concerning the convergence of the sequence $x_k$ to the solution under this hypothesis. Moreover, we show how to apply this this method to compute rigorously zeros for two-point boundary value problems of Neumann type. Finally, we apply it to a version of the Cahn-Hilliard equation.
△ Less
Submitted 15 May, 2020;
originally announced May 2020.
-
Rigorous Enclosures of Solutions of Neumann Boundary Value Problems
Authors:
Eduardo Ramos,
Victor Nolasco,
Marcio Gameiro
Abstract:
This paper is dedicated to the problem of isolating and validating zeros of non-linear two point boundary value problems. We present a method for such purpose based on the Newton-Kantorovich Theorem to rigorously enclose isolated zeros of two point boundary value problem with Neumann boundary conditions.
This paper is dedicated to the problem of isolating and validating zeros of non-linear two point boundary value problems. We present a method for such purpose based on the Newton-Kantorovich Theorem to rigorously enclose isolated zeros of two point boundary value problem with Neumann boundary conditions.
△ Less
Submitted 21 May, 2022; v1 submitted 6 May, 2020;
originally announced May 2020.
-
A Framework for the Numerical Computation and a Posteriori Verification of Invariant Objects of Evolution Equations
Authors:
Jordi-Lluís Figueras,
Marcio Gameiro,
Jean Philippe Lessard,
Rafael de la Llave
Abstract:
We develop a theoretical framework for computer-assisted proofs of the existence of invariant objects in semilinear PDEs. The invariant objects considered in this paper are equilibrium points, traveling waves, periodic orbits and invariant manifolds attached to fixed points or periodic orbits. The core of the study is writing down the invariance condition as a zero of an operator. These operators…
▽ More
We develop a theoretical framework for computer-assisted proofs of the existence of invariant objects in semilinear PDEs. The invariant objects considered in this paper are equilibrium points, traveling waves, periodic orbits and invariant manifolds attached to fixed points or periodic orbits. The core of the study is writing down the invariance condition as a zero of an operator. These operators are in general not continuous, so one needs to smooth them by means of preconditioners before classical fixed point theorems can be applied. We develop in detail all the aspects of how to work with these objects: how to precondition the equations, how to work with the nonlinear terms, which function spaces can be useful, and how to work with them in a computationally rigorous way. In two companion papers, we present two different implementations of the tools developed in this paper to study periodic orbits.
△ Less
Submitted 3 May, 2016;
originally announced May 2016.
-
Rigorous numerics for ill-posed PDEs: periodic orbits in the Boussinesq equation
Authors:
R. Castelli,
M. Gameiro,
J. -P. Lessard
Abstract:
In this paper, we develop computer-assisted techniques for the analysis of periodic orbits of ill-posed partial differential equations. As a case study, our proposed method is applied to the Boussinesq equation, which has been investigated extensively because of its role in the theory of shallow water waves. The idea is to use the symmetry of the solutions and a Newton-Kantorovich type argument (t…
▽ More
In this paper, we develop computer-assisted techniques for the analysis of periodic orbits of ill-posed partial differential equations. As a case study, our proposed method is applied to the Boussinesq equation, which has been investigated extensively because of its role in the theory of shallow water waves. The idea is to use the symmetry of the solutions and a Newton-Kantorovich type argument (the radii polynomial approach), to obtain rigorous proofs of existence of the periodic orbits in a weighted $\ell^1$ Banach space of space-time Fourier coefficients with geometric decay. We present several computer-assisted proofs of existence of periodic orbits at different parameter values.
△ Less
Submitted 29 September, 2015;
originally announced September 2015.
-
Continuation of Point Clouds via Persistence Diagrams
Authors:
Marcio Gameiro,
Yasuaki Hiraoka,
Ippei Obayashi
Abstract:
In this paper, we present a mathematical and algorithmic framework for the continuation of point clouds by persistence diagrams. A key property used in the method is that the persistence map, which assigns a persistence diagram to a point cloud, is differentiable. This allows us to apply the Newton-Raphson continuation method in this setting. Given an original point cloud $P$, its persistence diag…
▽ More
In this paper, we present a mathematical and algorithmic framework for the continuation of point clouds by persistence diagrams. A key property used in the method is that the persistence map, which assigns a persistence diagram to a point cloud, is differentiable. This allows us to apply the Newton-Raphson continuation method in this setting. Given an original point cloud $P$, its persistence diagram $D$, and a target persistence diagram $D'$, we gradually move from $D$ to $D'$, by successively computing intermediate point clouds until we finally find a point cloud $P'$ having $D'$ as its persistence diagram. Our method can be applied to a wide variety of situations in topological data analysis where it is necessary to solve an inverse problem, from persistence diagrams to point cloud data.
△ Less
Submitted 9 June, 2015;
originally announced June 2015.
