Closed-Form Robustness Bounds for Second-Order Pruning of Neural Controller Policies

TL;DR

This paper provides a rigorous mathematical analysis of how second-order pruning affects the robustness and stability of neural network controllers in nonlinear discrete-time systems, offering explicit bounds on pruning effects.

Contribution

It introduces the first closed-form robustness bounds for second-order pruning of neural controllers, linking network compression with control performance guarantees.

Findings

Derived explicit bounds on pruning-induced perturbations in neural controllers.

Provided a method to evaluate maximum pruning levels before control performance degrades.

Linked second-order network compression techniques with safety-critical control guarantees.

Abstract

Deep neural policies have unlocked agile flight for quadcopters, adaptive grasping for manipulators, and reliable navigation for ground robots, yet their millions of weights conflict with the tight memory and real-time constraints of embedded microcontrollers. Second-order pruning methods, such as Optimal Brain Damage (OBD) and its variants, including Optimal Brain Surgeon (OBS) and the recent SparseGPT, compress networks in a single pass by leveraging the local Hessian, achieving far higher sparsity than magnitude thresholding. Despite their success in vision and language, the consequences of such weight removal on closed-loop stability, tracking accuracy, and safety have remained unclear. We present the first mathematically rigorous robustness analysis of second-order pruning in nonlinear discrete-time control. The system evolves under a continuous transition map, while the controller is an $L$-layer multilayer perceptron with ReLU-type activations that are globally 1-Lipschitz. Pruning the weight matrix of layer $k$ replaces $W_k$ with $W_k+\delta W_k$, producing the perturbed parameter vector $\widehat{\Theta}=\Theta+\delta\Theta$ and the pruned policy $\pi(\cdot;\widehat{\Theta})$. For every input state $s\in X$ we derive the closed-form inequality $ \|\pi(s;\Theta)-\pi(s;\widehat{\Theta})\|_2 \le C_k(s)\,\|\delta W_k\|_2, $ where the constant $C_k(s)$ depends only on unpruned spectral norms and biases, and can be evaluated in closed form from a single forward pass. The derived bounds specify, prior to field deployment, the maximal admissible pruning magnitude compatible with a prescribed control-error threshold. By linking second-order network compression with closed-loop performance guarantees, our work narrows a crucial gap between modern deep-learning tooling and the robustness demands of safety-critical autonomous systems.