{
 "artifacts": null,
 "category": "dynamics",
 "description": "Run several heterogeneous recurrent or state-space network copies and couple their hidden states through a directed hypergraph with proportional and integral feedback. The proportional term contracts disagreement, while the integral state rejects persistent replica-specific biases that ordinary consensus coupling can only bound. This creates a controllable synchronization-versus-divergence transition rather than an unstructured regularization coefficient.",
 "formulas_latex": [
  "$$\\dot{x}_i=f_i(x_i,\\mu_i)-k_P\\sum_{j=1}^{N}L^P_{ij}g(x_j)-z_i,\\qquad \\dot{z}_i=k_I\\sum_{j=1}^{N}L^I_{ij}g(x_j).$$",
  "$$\\mathcal{M}(\\lambda_P,\\lambda_I)=\\begin{bmatrix}J_xf(\\bar{x},\\bar{\\mu})-\\lambda_PJ_xg(0)\u0026-I_n\\\\ \\lambda_IJ_xg(0)\u00260_n\\end{bmatrix},\\qquad \\chi_{\\max}(\\lambda_P,\\lambda_I)\u003c0.$$",
  "$$\\rho\\!\\left(I_{2n}+h\\mathcal{M}(\\lambda_P,\\lambda_I)\\right)\u003c1\\quad\\text{for all transverse modes}.$$",
  "$$\\delta\\dot{x}=\\left[J_xf-\\lambda_PJ_xg(0)\\right]\\delta x-\\delta z+J_\\mu f\\,\\delta\\mu,\\qquad \\delta\\dot{z}=\\lambda_IJ_xg(0)\\delta x.$$"
 ],
 "id": 3144,
 "implementation": "1. Integration point: create N recurrent or state-space network copies with the same task architecture but deliberately heterogeneous parameters, such as different low-rank perturbations, quantization noise, or expert-specific biases. At every sequence step, construct a directed hypergraph over copies; ordinary directed edges are the MVP, while a hyperedge can aggregate the mean state of several tail copies and inject it into several head copies. Couple a selected hidden state h_i or a low-dimensional projection g(h_i), not the full parameter vector. 2. Pseudocode: initialize h_i and integral states z_i to zero; for each time step compute h_i_next = f_i(h_i,input); compute yP_i = sum_j Lp[i,j] g(h_j) and yI_i = sum_j Li[i,j] g(h_j); update z_i = z_i + h*kI*yI_i; update h_i = h_i + h*(f_i(h_i,input) - kP*yP_i - z_i); compute the task loss from the average or one designated synchronized copy and backpropagate through the coupled rollout. Start with g(h) equal to a linear projection and zero-row-sum Laplacians. 3. Computed from the paper: the transverse matrix and the requirement that its maximum Lyapunov exponent be negative. Estimated empirically: J_x f by Jacobian-vector products along validation sequences, Laplacian eigenvalues numerically, and the maximum Lyapunov exponent by QR-based tangent propagation. 4. First cheap experiment: compare five GRU copies on sequential MNIST or permuted-MNIST, with fixed per-copy hidden biases, against uncoupled training and proportional-only coupling. Sweep kP, kI, and Euler step h. Measure pairwise hidden disagreement and validation accuracy. The predicted signature is a sharp stability boundary: synchronization should persist only where the largest estimated transverse exponent is negative, and the observed divergence boundary should agree with chi_max = 0 within roughly 20%. With a constant injected bias, proportional-only coupling should leave nonzero steady disagreement, whereas PI coupling should drive disagreement toward zero; increasing kI should remove the residual bias until a discrete-time instability boundary is crossed.",
 "math_summary": "Let x_i in R^n be the hidden state of replica i, f_i its vector field, g the coupling output, L^P and L^I directed hypergraph Laplacians, k_P and k_I their coupling gains, and z_i in R^n an integral controller state. Use dx_i/dt = f_i(x_i,mu_i) - k_P sum_j L^P_ij g(x_j) - z_i and dz_i/dt = k_I sum_j L^I_ij g(x_j). The synchronization manifold is x_1 = ... = x_N. For a transverse Laplacian mode with eigenvalues lambda_P and lambda_I, linearization around the synchronized trajectory gives M(lambda_P,lambda_I) = [[J_x f(x_bar,mu_bar) - lambda_P J_x g(0), -I_n], [lambda_I J_x g(0), 0_n]]. Here J_x f is the hidden-state Jacobian, J_x g is the coupling-output Jacobian, and I_n is the n-dimensional identity. The paper's condition is that the maximum Lyapunov exponent of every transverse M-mode is negative. For a discrete Euler implementation with step h, a local stability test is rho(I_2n + h M(lambda_P,lambda_I)) \u003c 1 for every non-consensus mode. Exact synchronization requires the mismatch contribution J_mu f(x_bar,mu_bar) delta_mu to be constant in the transverse dynamics; the integral state can then cancel it.",
 "math_tags": [
  "control-theory",
  "dynamical-systems",
  "spectral-theory",
  "linear-algebra"
 ],
 "ml_areas": [
  "rnn",
  "ssm",
  "training-dynamics",
  "optimizer"
 ],
 "paper": {
  "arxiv_id": "2609.03698",
  "arxiv_url": "https://arxiv.org/abs/2609.03698",
  "summary_what_math_gives_to_ml": "The paper provides a transferable master-stability mechanism for synchronizing heterogeneous nonlinear agents using dynamic proportional-integral hypergraph coupling. Its key result is that constant parameter-mismatch disturbances can be exactly rejected by an integral coupling state, while local synchronization is guaranteed when every transverse mode has negative maximum Lyapunov exponent. A practical neural-network adaptation is to couple parallel recurrent or state-space network copies through a directed higher-order interaction graph, adding an integral consensus controller that cancels persistent replica-specific biases. The sharp testable signature is a synchronization boundary where the transverse Lyapunov exponent crosses zero, or, after discretization, where the spectral radius of the transverse update crosses one.",
  "title": "Synchronization of directed hypergraphs with heterogeneities via dynamic coupling",
  "year": "2026"
 },
 "ratings": {
  "difficulty": 6,
  "novelty": 8,
  "usefulness": 7
 },
 "solves": [
  "stability",
  "accuracy",
  "generalization"
 ],
 "title": "Integral Master-Stability Coupling for Heterogeneous RNN Copies",
 "url": "https://synthcore.org/idea/3144/integral-master-stability-coupling-for-heterogeneous-rnn-cop",
 "verification": {
  "peer_reviewed": false,
  "status": "unverified",
  "status_label": "Unverified",
  "verdict_source": "deterministic test code (paired-seed permutation statistics)",
  "verification_axes": {
   "benchmark_mechanism": {
    "confirmed": null,
    "tested": false
   },
   "practical_benchmark": {
    "beats_baseline": null,
    "tested": false
   },
   "toy_mechanism_gate": {
    "confirmed": null,
    "tested": false
   }
  }
 }
}