{
 "artifacts": null,
 "category": "architecture",
 "description": "Build actuator or parameter constraints directly into the neural controller using a differentiable hard-saturation map rather than penalizing violations after the fact. This makes the Lyapunov certificate apply to the actual bounded controller and prevents training from exploiting unrealistically large actions.",
 "formulas_latex": [
  "$$u_{i}(x)=u_{i,\\min}+(u_{i,\\max}-u_{i,\\min})\\sigma(a_{\\theta,i}(x)),\\qquad \\sigma(s)=\\frac{1}{1+e^{-s}},$$",
  "$$u(x)=u_{\\max}\\tanh(a_{\\theta}(x)),\\qquad \\|u(x)\\|_{\\infty}\u003cu_{\\max},$$",
  "$$V_{\\psi}\\!\\left(\\hat f_{\\phi}(x,u(x))\\right)-V_{\\psi}(x)\\leq-c_{2}\\|x-x^{\\star}\\|_{2}^{2},$$",
  "$$r_{\\mathrm{cert}}(u_{\\max})=\\sup\\{r\u003e0:\\Delta V(x)\\leq-c_{2}\\|x-x^{\\star}\\|_{2}^{2}\\ \\forall\\|x-x^{\\star}\\|_{2}\\leq r\\}.$$"
 ],
 "id": 2825,
 "implementation": "1. Integration point: replace the final action head in an RL policy, learned controller, or neural ODE control input with a bounded sigmoid or tanh map. Do not merely clip actions outside the computational graph: the saturated action must be passed into the learned dynamics and Lyapunov calculations. 2. Pseudocode: compute logits $a=\\mathrm{Policy}_{\\theta}(x)$; map them to $u$ using the displayed bound-preserving formula; predict $x^{+}=\\hat f_{\\phi}(x,u)$; evaluate the task loss and Lyapunov decrease penalty; backpropagate through the saturation map. Initialize the final-layer scale so that most actions lie in the nonsaturated central region, then gradually impose the actual actuator bounds. At deployment, log the saturation fraction and Lyapunov margin. 3. The hard-bound guarantee is exact and analytical. The effect of saturation on stability is not assumed: estimate the decrease margin and certified radius numerically or with interval verification using the actual bounded action. 4. First cheap experiment: train identical MLP policies on a two-dimensional nonlinear stabilization task with action limits $u_{\\max}\\in\\{0.25,0.5,1,2\\}$, comparing unconstrained output plus post-hoc clipping against the hard-saturated architecture. For each limit, estimate $r_{\\mathrm{cert}}$ and the fraction of states at which the Lyapunov decrease inequality holds. The predicted signature is monotone shrinkage of $r_{\\mathrm{cert}}$ as $u_{\\max}$ decreases, with a qualitative transition when saturation prevents the required control authority; hard saturation should have no action-bound violations, whereas post-hoc clipping can show a certificate mismatch because the verified action differs from the trained action.",
 "math_summary": "For componentwise input bounds $u_{i,\\min}\u003cu_{i,\\max}$, let $a_{\\theta}(x)$ be an unconstrained neural output and define $u_{i}(x)=u_{i,\\min}+(u_{i,\\max}-u_{i,\\min})\\sigma(a_{\\theta,i}(x))$, where $\\sigma(s)=1/(1+e^{-s})$. Equivalently, a symmetric bound is $u=u_{\\max}\\tanh(a_{\\theta}(x))$. The controller always satisfies $u\\in\\mathcal{U}$, so the closed-loop transition used in the Lyapunov condition is the saturated transition $\\hat f(x,u_{\\mathrm{sat}})$ rather than the unconstrained one. Saturation can reduce local control authority; therefore the certificate's decrease margin must be recomputed after saturation and the feasible region is expected to shrink when bounds become tight.",
 "math_tags": [
  "control-theory",
  "dynamical-systems",
  "optimization",
  "linear-algebra"
 ],
 "ml_areas": [
  "rl",
  "mlp",
  "training-dynamics",
  "regularization"
 ],
 "paper": {
  "arxiv_id": "2608.29303",
  "arxiv_url": "https://arxiv.org/abs/2608.29303",
  "summary_what_math_gives_to_ml": "The paper offers a constructive pipeline for learning feedback policies for unknown nonlinear dynamics: identify a model from offline data, jointly synthesize a neural controller and neural Lyapunov function, enforce inputs with hard saturation, and certify the result using SMT plus local Lyapunov analysis. The transferable asset is a closed-loop certificate that turns policy learning into a falsifiable decrease-condition problem. A strong neural-network transfer is to train policies or recurrent state updates together with an energy function whose one-step change is negative on a specified region, while explicitly accounting for model-identification error. The most useful experimental signature is a certified region of attraction and a sharp failure boundary when the Lyapunov decrease margin becomes nonpositive.",
  "title": "Learning neural controllers for nonlinear systems from data",
  "year": "2026"
 },
 "ratings": {
  "difficulty": 3,
  "novelty": 4,
  "usefulness": 6
 },
 "solves": [
  "stability",
  "accuracy"
 ],
 "title": "Hard-Saturated Neural Feedback",
 "url": "https://synthcore.org/idea/2825/hard-saturated-neural-feedback",
 "verification": {
  "peer_reviewed": false,
  "status": "unverified",
  "status_label": "Unverified",
  "verdict_source": "deterministic test code (paired-seed permutation statistics)"
 }
}