---
title: Dynamic Local Convergence
url: https://www.emergentmind.com/topics/dynamic-local-convergence
type: topic
---

# Dynamic Local Convergence

Searching arXiv for recent papers using the term “dynamic local convergence” and closely related formulations.
Dynamic local convergence is a context-dependent research term used in several mathematical, algorithmic, and applied domains to denote convergence behavior governed by local structure in a time-evolving or iterative system. Across the literature, the phrase refers to distinct but structurally related phenomena: local linearization around equilibria in nonconvex optimization dynamics [2305.08277]; nonlocal-to-local limits in evolutionary PDEs with time dependence [2402.12145]; consensus and optimizer convergence of local agent states in distributed optimization [2309.12235]; convergence of local microscopic distributions to evolving equilibrium states in interacting particle systems [1804.06148]; maintenance of locally valid solutions under adversarial updates in algorithmic Lovász Local Lemma settings [2604.20836]; and convergence of epidemic processes on dynamic random graphs to a process on a dynamic local limit object [2501.09623]. This diversity suggests that “dynamic local convergence” is not a single formalism, but a recurring paradigm in which local geometry, locality of interaction, or local state variables determine asymptotic behavior under dynamics.

## 1. Dynamic local convergence as a cross-disciplinary concept

In optimization and learning, the term often denotes convergence of iterates near a stationary point, with the local rate controlled by linearized dynamics, curvature, regularization, and algorithmic step sizes. In the GAN setting, local convergence of gradient descent–ascent is characterized through the Jacobian spectrum of a linearization around equilibrium, with convergence, oscillation, and divergence separated by explicit phase boundaries [2305.08277]. In primal–dual and splitting methods, a related pattern appears: iterates first identify active smooth manifolds and then enter a local linear regime governed by a linearized operator restricted to tangent spaces [1705.01926, 1606.02116].

In distributed and federated optimization, the phrase emphasizes the convergence of each agent’s local variable. In distributed conjugate-gradient tracking, local convergence means that each agent’s local variable converges to the global optimizer, $\|x_i^k-x^*\|\to 0$, under consensus and vanishing mean gradient conditions [2309.12235]. In federated learning with periodic averaging, the relevant local dynamics are those of client drift and synchronization; convergence depends on gradient diversity, aggregation frequency, and network topology, and the analysis identifies when local updates plus averaging recover rates matching homogeneous settings [1910.14425].

In stochastic particle systems and random graphs, the term shifts from iterates to local laws. For the asymmetric zero-range process with site disorder, dynamic local convergence means that under hydrodynamic scaling the local microscopic distribution around a macroscopic point converges to the quenched Gibbs equilibrium determined by the evolving macroscopic density, or to the critical measure when local mass loss occurs [1804.06148]. For SIR epidemics on dynamic random graphs, the paper formalizes a stronger local time-marked union convergence because snapshotwise local weak convergence does not capture time-aggregated infection paths [2501.09623].

A plausible implication is that the unifying core of the term is not a common theorem template, but a common viewpoint: one studies a time-dependent global system by proving convergence of a local object, such as a neighborhood, an agent state, an active manifold, or a local law.

## 2. Local linearization, spectra, and phase behavior in optimization dynamics

A canonical instance is the analysis of GAN training with kernel discriminators. The model is posed as the min–max problem
\[
\min_{\theta}\max_{\phi}\;\mathcal{L}(\theta,\phi),
\]
with generated point locations as generator parameters and an RKHS discriminator regularized by $\lambda>0$ [2305.08277]. Under an isolated points model, the dynamics decouple locally around each true point, and the update map can be linearized near equilibrium. The Jacobian takes the block form
\[
J=\begin{bmatrix} I-\alpha H_{\theta\theta} & -\alpha H_{\theta\phi}\\ \beta H_{\phi\theta} & I+\beta H_{\phi\phi} \end{bmatrix},
\]
with $\alpha=\eta_g$ and $\beta=\eta_d$ [2305.08277]. Local linear convergence occurs when the spectral radius satisfies $\rho(J)<1$, oscillations arise when complex eigenvalues have modulus near $1$, and divergence occurs when $\rho(J)>1$ [2305.08277].

The resulting eigenvalue formulas expose explicit dependence on regularization $\lambda$, kernel bandwidth $\sigma$, the step-size ratio $\mu=\eta_g/\eta_d$, and local mass mismatch $\Delta_i$ [2305.08277]. The paper gives a simple sufficient condition for local stability,
\[
\eta_d<\frac{2}{\lambda},\qquad \eta_g<\lambda\,\sigma^2,
\]
whenever $\Delta_i>0$ [2305.08277]. It also identifies an oscillatory regime defined by $m^2<c$, and a saturation regime in which decreasing $\sigma$ no longer improves the rate because the dominant eigenvalue is controlled by regularization [2305.08277]. This is an explicit example of dynamic local convergence as a phase-structured property of a linearized non-Markovian system.

A comparable two-stage pattern appears in nonsmooth convex splitting. For primal–dual splitting, the local theory first proves finite identification of primal and dual smooth manifolds and then proves local linear convergence characterized by the spectral radius of a linearized operator on the active manifolds [1705.01926]. For Douglas–Rachford and ADMM, the same geometry yields finite manifold identification followed by linear convergence, and in the locally polyhedral case the optimal contraction factor is given by the cosine of the Friedrichs angle between the tangent spaces of the identified submanifolds [1606.02116]. These results make locality geometric: convergence is dictated by tangent spaces, active sets, and restricted curvature rather than by global convexity alone.

Related local analyses also appear in optimal control and manifold saddle dynamics. Generalized Gauss–Newton multiple shooting, single shooting, and differential dynamic programming share the same local $Q$-linear rate when a GGN Hessian approximation is used, because their iteration Jacobians agree at the solution [2301.04047]. On manifolds, discrete constrained saddle dynamics achieve local linear convergence with rate depending on the condition number of the Riemannian Hessian, while a momentum variant improves the dependence from $\kappa$ to $\sqrt{\kappa}$ in the rate bound [2601.22341]. Continuous-time weakly contracting but locally strongly contracting systems exhibit a distinct linear–exponential convergence profile, with linear decay outside the strongly contracting neighborhood and exponential decay once the trajectory enters it [2403.07572].

## 3. Distributed, federated, and hierarchical formulations

In distributed optimization, dynamic local convergence often refers to convergence of each local state under communication and local computation. In the distributed conjugate-gradient method with conjugate direction tracking, each agent maintains a local variable $x_i^k$, a local conjugate direction $s_i^k$, and a dynamic consensus tracker $z_i^k$ for the average direction [2309.12235]. The key update
\[
z_i^{k+1} = \sum_{j\in N_i\cup\{i\}} W_{ij}\big[z_j^k + (s_j^{k+1}-s_j^k)\big]
\]
ensures mean preservation of the tracked direction under a doubly stochastic mixing matrix [2309.12235]. Under bounded $\alpha_i^k$ and $\beta_i^k$ and contraction conditions $\lambda_s<1$ and $\lambda_z<1$, the paper proves consensus of $x$, $s$, and $z$, vanishing mean gradient, and convergence of objective values to $f(x^*)$ without diminishing step sizes [2309.12235]. Under strong convexity, this implies $\|x_i^k-x^*\|\to 0$ for all agents [2309.12235].

In federated learning, local descent with periodic averaging is analyzed through perturbation of the averaged iterate:
\[
\bar w^{(t+1)}=\bar w^{(t)}-\eta_t \nabla f(\bar w^{(t)})+\eta_t \Delta^{(t)},
\]
where $\Delta^{(t)}$ captures heterogeneity, stochastic variance, and client drift [1910.14425]. The heterogeneity measure is the weighted gradient diversity
\[
\Lambda(w,q)=\frac{\sum_{j=1}^p q_j \|\nabla f_j(w)\|^2}{\|\sum_{j=1}^p q_j \nabla f_j(w)\|^2}\le \lambda
\]
[1910.14425]. The analysis shows that periodic averaging and suitable choices of $\eta$, $E$, and $K$ control the drift and recover rates matching the best-known homogeneous setting in general nonconvex and PL cases [1910.14425]. This is a local convergence theory in the sense that client-side local models synchronize frequently enough for the global descent term to dominate the local deviations.

Hierarchical SGD adds another layer: local aggregation at intermediate servers partitions heterogeneity into “upward” and “downward” divergences [2010.12998]. The global divergence decomposes exactly into across-group and within-group components, and the worst-case bound for hierarchical SGD lies between the bounds of two single-level local SGD schemes, one using the local aggregation period and the other using the global aggregation period [2010.12998]. The paper terms this the “sandwich behavior” [2010.12998]. This suggests that dynamic local convergence in hierarchical systems can be understood as a balance between local contraction inside groups and slower correction of global drift.

A different distributed example arises in solving discrete-time Lyapunov equations over time-varying graphs. There, local convergence means that each agent’s local iterate converges to the common DTLE solution despite changing neighbor sets, under uncoordinated constant stepsizes and uniform joint connectivity [1904.13067]. The analysis decomposes each agent’s error into consensus error and optimization error, both of which decay linearly under the stated assumptions [1904.13067].

## 4. Dynamic local convergence of local laws, neighborhoods, and limits

In interacting particle systems, dynamic local convergence concerns the evolution of local probability laws rather than finite-dimensional iterates. For the asymmetric zero-range process with site disorder, the hydrodynamic density solves a scalar conservation law with flux
\[
f(\rho)=
\begin{cases}
(p-q)\,\beta(\rho), & \rho<\rho_c,\\
(p-q)\,\phi_c, & \rho\ge \rho_c,
\end{cases}
\]
and the microscopic process exhibits quenched strong local equilibrium when $\rho(t,u)<\rho_c$ [1804.06148]. At supercritical densities, no invariant Gibbs product measure exists with that density, and the local law converges instead to the critical product equilibrium, producing dynamic local loss of mass [1804.06148]. The local object is the law seen in a bounded microscopic window around the macroscopic point $(t,u)$.

In phase-field PDEs, “dynamic local convergence” refers to nonlocal-to-local passage in a time-dependent system with inertia [2402.12145]. The nonlocal phase equation
\[
(\varphi_\varepsilon)_{tt}+(\varphi_\varepsilon)_t+B_\varepsilon\varphi_\varepsilon+\beta(\varphi_\varepsilon)+\pi(\varphi_\varepsilon)=\theta_\varepsilon
\]
converges, as $\varepsilon\to 0$, to the local equation
\[
\varphi_{tt}+\varphi_t-\Delta\varphi+\beta(\varphi)+\pi(\varphi)=\theta
\]
in strong and weak time-space topologies [2402.12145]. The convergence mechanism relies on operator convergence $B_\varepsilon u_\varepsilon \rightharpoonup Bu$ in $V^*$ under energy bounds, compactness, and a precise normalization of the kernels [2402.12145]. Here locality means replacement of a convolution-type interaction by the Neumann Laplacian in the limit.

Dynamic random graph theory provides another notion. For SIR epidemics on evolving graphs, the required convergence is not merely dynamic local weak convergence of snapshots but local time-marked union convergence in probability [2501.09623]. The dynamic graph is converted into a union graph over $[0,T]$ whose edges are marked by ON/OFF activation times, and convergence is imposed in a metric on rooted marked graphs [2501.09623]. This stronger notion is necessary because infection may travel along time-aggregated paths that never appear inside a single connected component at any fixed time [2501.09623]. Under this notion, the empirical epidemic proportions converge at fixed time $t$ to deterministic limits determined by the epidemic on the dynamic local limit graph [2501.09623].

Dynamic random intersection graphs develop a related process-level local convergence theory. The time-evolving graph is represented as a marked bipartite branching-process limit, and finite-dimensional convergence plus tightness establish dynamic local weak convergence of rooted neighborhoods [2308.15629]. The framework then transfers to time-dependent degree distributions and giant-component membership processes [2308.15629]. In both [2501.09623] and [2308.15629], local convergence is explicitly a statement about neighborhoods in a dynamic Polish-space setting rather than about optimization errors.

## 5. Fully dynamic algorithms and online adaptation

Some uses of the term concern maintaining locally valid structure under changing problem instances. In the dynamic Lovász Local Lemma setting, a search problem is specified by a finite state space $\Omega$ and a set of flaws $F=\{F_1,\dots,F_m\}$; an adaptive or even clairvoyant adversary inserts or deletes flaws over time [2604.20836]. Dynamic local convergence means that the same local search, resampling, or backtracking procedures used in the static setting continue to converge rapidly after each update, with total local fixing steps
\[
O_{\psi,\varepsilon}(q+m+\log|\Omega|)
\]
over $q$ updates and failure probability at most $2^{-q}$ [2604.20836]. The central condition is the AIS-type convergence criterion
\[
\frac{1}{\psi_i}\sum_{S\subseteq \Gamma_i}\gamma_i^S \prod_{j\in S}\psi_j \le 1-\varepsilon,
\]
which controls witness-forest growth under dynamic updates [2604.20836]. This is a notion of dynamic local convergence because the locality lies in flaw-specific fixing operations and dependency neighborhoods, while the dynamic aspect is adversarial online modification.

Online optimization for recurrent and adaptive algorithms offers another interpretation. For RTRL, NoBackTrack, UORO, RMSProp, online natural gradient, and Adam with $\beta^2\to 1$, local convergence is analyzed in an ergodic online setting where the state evolves according to a parameterized dynamical system and the parameter is updated while the system runs [2005.05645]. The general limiting ODE is
\[
\dot{\theta}=-U(\theta),
\]
and local convergence holds when the averaged Jacobian at the target parameter is positive-stable and the target trajectory is locally stable [2005.05645]. The framework departs from standard SGD by using empirical time averages rather than i.i.d. expectations, which changes admissible stepsize ranges and permits larger rates under cycling or reshuffling [2005.05645].

Dynamic topology optimization introduces yet another usage. There, local convergence refers to the tendency of gradient-based optimizers to settle in low-performing stationary points shaped by resonances, antiresonances, and symmetry, often producing stiff, mass-driven designs instead of mechanism-rich solutions [2509.24667]. The paper organizes mitigation strategies into exclusion, frequency shift, and relaxation approaches, and quantifies how these alter the probability of reaching high-performing optima over frequency bands [2509.24667]. This suggests a broader reading: dynamic local convergence can also denote undesirable convergence to nearby poor minima in highly nonconvex dynamic design landscapes.

## 6. Common structure, distinctions, and limitations

A common structural pattern recurs across these literatures. First, a local object is identified: an equilibrium neighborhood with explicit Jacobian in GAN training [2305.08277], an active manifold in splitting methods [1705.01926, 1606.02116], a tangent-space Hessian on a manifold [2601.22341], an agent-local state in distributed optimization [2309.12235, 1904.13067], a microscopic window in stochastic particle systems [1804.06148], or a rooted marked neighborhood in dynamic graph models [2501.09623, 2308.15629]. Second, convergence is controlled by a local descriptor: spectral radius, contraction factor, Friedrichs angle, condition number, gradient diversity, or operator limit. Third, the theory often separates a global or transient phase from a local asymptotic phase. Examples include manifold identification followed by linear convergence [1705.01926, 1606.02116], linear-then-exponential decay in weakly contracting continuous dynamics [2403.07572], and local neighborhood approximation followed by convergence of epidemic observables [2501.09623].

Important distinctions remain. In some papers, “local” means spatial or graph-theoretic locality, as in random graphs and particle systems [2501.09623, 1804.06148]. In others, it means local stability near a stationary point or minimizer [2305.08277, 2403.07572, 2301.04047]. In distributed optimization, it refers to agent-local copies or local updates [2309.12235, 1910.14425, 2010.12998]. In nonlocal PDE limits, it refers to passage from nonlocal interaction kernels to local differential operators [2402.12145]. Because these meanings are not interchangeable, the term should generally be interpreted within its disciplinary framework rather than as a single universal definition.

The limitations are likewise domain-specific. Local linearization results do not imply global convergence and often rely on isolated equilibria, strict complementarity, or exact eigenspaces [2305.08277, 1705.01926, 2601.22341]. Dynamic random graph convergence may require time-marked union convergence rather than weaker snapshotwise notions [2501.09623]. Distributed exact convergence can require strong convexity, spectral gap conditions, or bounds on heterogeneity [2309.12235, 1910.14425, 2010.12998]. Nonlocal-to-local PDE results may provide only qualitative subsequential convergence without rates [2402.12145]. Fully dynamic LLL guarantees hold only while an LLL criterion with slack remains satisfied after each update [2604.20836].

This suggests that “dynamic local convergence” is best understood as a family of locality-based asymptotic principles. In each instance, dynamics are governed not by the full global state space at once, but by a local structure whose evolution is tractable and whose asymptotic behavior determines the phenomenon of interest.

Source: https://www.emergentmind.com/topics/dynamic-local-convergence