aiqbench analysis
QuantKAN establishes the first unified framework for quantizing Kolmogorov-Arnold Networks by adapting modern techniques like DSQ and GPTQ to their unique spline-based parameters. The analysis reveals that while low-bit deployment can significantly reduce energy consumption on hardware accelerators, specific architectures suffer from distinct failure modes depending on whether base or basis functions dominate the representation. This work provides essential benchmarks for deploying KANs in resource-constrained environments without sacrificing model expressivity.
QuantKAN: A Unified Quantization Framework for Kolmogorov Arnold Networks
arXiv:2511.18689v3 Announce Type: replace Abstract: Kolmogorov--Arnold Networks (KANs) replace linear weights with spline-based functions, offering strong expressivity but posing challenges for low-precision deployment due to heterogeneous parameter distributions. We introduce QuantKAN, the first unified framework for quantization-aware training (QAT) and post-training quantization (PTQ) of KANs. The framework employs branch-aware quantizers for base and spline parameters and extends modern QAT and PTQ methods to spline-based layers across EfficientKAN, FastKAN, PyKAN, and KAGN. Experiments on MNIST, CIFAR-10/100, TinyImageNet, and ImageNet provide the first unified QAT/PTQ KAN benchmarks and show that DSQ is the most robust QAT method at aggressive low-bit settings, while GPTQ is the strongest PTQ method at moderate precision. Sensitivity analyses reveal architecture-specific failure modes: spline/basis parameters dominate in FastKAN, while base or scaling parameters dominate in EfficientKAN, GRAM, and PyKAN. Vivado HLS estimates on a Xilinx UltraScale+ device further suggest up to 3.32$\times$ throughput and 7.7$\times$ lower estimated dynamic energy per inference under W4A4, exposing a residual \emph{basis-evaluation tax} that motivates basis-aware microarchitecture. QuantKAN is available at https://github.com/OSU-STARLAB/QuantKAN/.
Abstract:Kolmogorov--Arnold Networks (KANs) replace linear weights with spline-based functions, offering strong expressivity but posing challenges for low-precision deployment due to heterogeneous parameter distributions. We introduce QuantKAN, the first unified framework for quantization-aware training (QAT) and post-training quantization (PTQ) of KANs. The framework employs branch-aware quantizers for base and spline parameters and extends modern QAT and PTQ methods to spline-based layers across EfficientKAN, FastKAN, PyKAN, and KAGN. Experiments on MNIST, CIFAR-10/100, TinyImageNet, and ImageNet provide the first unified QAT/PTQ KAN benchmarks and show that DSQ is the most robust QAT method at aggressive low-bit settings, while GPTQ is the strongest PTQ method at moderate precision. Sensitivity analyses reveal architecture-specific failure modes: spline/basis parameters dominate in FastKAN, while base or scaling parameters dominate in EfficientKAN, GRAM, and PyKAN. Vivado HLS estimates on a Xilinx UltraScale+ device further suggest up to 3.32$\times$ throughput and 7.7$\times$ lower estimated dynamic energy per inference under W4A4, exposing a residual \emph{basis-evaluation tax} that motivates basis-aware microarchitecture. QuantKAN is available at this https URL.
Machine Learning (cs.LG)
arXiv:2511.18689 [cs.LG]
(or arXiv:2511.18689v3 [cs.LG] for this version)
Full text under the publisher's open licence · 177 words
Related coverage
FuncCode: Compressing Kolmogorov--Arnold Networks in Function Space with Hardware-Aware Quantization
arXiv · Sep 23, 2026
On aggregation-quantization permutability problem for discrete-time Markov chains
arXiv · Oct 28, 2026
Instabilities in a Non-KAM System via Information Scrambling: A Note
arXiv · Jun 12, 2026
Quantizing With Randomized Hadamard Transforms: Efficient Heuristic Now Proven
arXiv · Nov 4, 2026