# An Embedded RISC-V Evaluation of Kolmogorov--Arnold Networks in Hard-Constrained Recurrent Physics-Informed Models

arXiv: https://arxiv.org/abs/2608.00737 · HTML: https://mlsanity.com/paper/2608.00737/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 14 of its 20 references that are in the map.
- First posted on arXiv 1 Aug 2026.

## Builds on (most cited first, 14 of 14)

- [Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference](https://mlsanity.com/paper/1712.05877/) (2017)
- [KAN: Kolmogorov-Arnold Networks](https://mlsanity.com/paper/2404.19756/) (2024)
- [Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning](https://mlsanity.com/paper/2108.06266/) (2021)
- [A comprehensive and FAIR comparison between MLP and KAN representations for differential equations and operator networks](https://mlsanity.com/paper/2406.02917/) (2024)
- [DeepOKAN: Deep Operator Network Based on Kolmogorov Arnold Networks for Mechanics Problems](https://mlsanity.com/paper/2405.19143/) (2024)
- [KAN-ODEs: Kolmogorov-Arnold Network Ordinary Differential Equations for Learning Dynamical Systems and Hidden Physics](https://mlsanity.com/paper/2407.04192/) (2024)
- [Physics-Informed Neural Networks with Hard Linear Equality Constraints](https://mlsanity.com/paper/2402.07251/) (2024)
- [A Practitioner's Guide to Kolmogorov-Arnold Networks](https://mlsanity.com/paper/2510.25781/) (2025)
- [Hardware Acceleration of Kolmogorov-Arnold Network (KAN) for Lightweight Edge Inference](https://mlsanity.com/paper/2409.11418/) (2024)
- [Structured Kolmogorov-Arnold Neural ODEs for Interpretable Learning and Symbolic Discovery of Nonlinear Dynamics](https://mlsanity.com/paper/2506.18339/) (2025)
- [KANtize: Exploring Low-bit Quantization of Kolmogorov-Arnold Networks for Efficient Inference](https://mlsanity.com/paper/2603.17230/) (2026)
- [Hard-Constrained Neural Networks with Physics-Embedded Architecture for Residual Dynamics Learning and Invariant Enforcement in Cyber-Physical Systems](https://mlsanity.com/paper/2511.23307/) (2025)
- [Empirical Stability Analysis of Kolmogorov-Arnold Networks in Hard-Constrained Recurrent Physics-Informed Discovery](https://mlsanity.com/paper/2602.09988/) (2026)
- [Hardware-Oriented Inference Complexity of Kolmogorov-Arnold Networks](https://mlsanity.com/paper/2604.03345/) (2026)

## Abstract (arXiv)

Hard-constrained recurrent physics-informed networks (HRPINNs) embed known dynamics inside a recurrent numerical integrator and restrict a neural branch to learning only the residual dynamics that the first-principles model does not capture. Kolmogorov--Arnold Networks (KANs) have been proposed as parameter-efficient replacements for multilayer perceptrons (MLPs) in such residual branches, but their learnable B-spline activations follow a markedly different execution profile. Building on prior work that characterized when a vanilla B-spline KAN matches or underperforms an MLP as an HRPINN residual branch in discovery accuracy, this paper asks whether that parameter efficiency survives deployment. Using identical trained weights, we measured execution latency, energy per integration step, and dependability under post-training quantization in the closed recurrent loop on a RISC-V RV64GC platform without vector extensions (StarFive VisionFive~2, SiFive U74). For the two accuracy-comparable pairs, the KAN residual branch executed 13.5× and 8.0× slower and consumed 11.3× and 5.6× more energy per integration step (3.7,μJ against 0.33,μJ for the smallest pair); across all four parameter-matched size tiers the ranges are 4.7×--14.5× and 4.7×--18.7×. Under INT8 quantization, KAN trajectories diverged up to 43× earlier than matched MLPs; the damage traces to weight quantization, not to input-side knot-interval misassignment. These results indicate that the parameter efficiency reported for KANs does not transfer to deployment cost on scalar embedded cores, and that an MLP residual branch is the more dependable default for embedded HRPINN deployment unless specific quantization co-design is used.