cs.ARFeb 26, 2026

BiKA: Kolmogorov-Arnold-Network-inspired Ultra Lightweight Neural Network Hardware Accelerator

Authors: Yuhao Liu, Salim Ullah, Akash Kumar

Organizations: Ruhr University Bochum, Germany · Dresden University of Technology, Germany · Center for Scalable Data Analytics and Artificial Intelligence (ScaDS.AI Dresden/Leipzig), Germany

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Lightweight neural network accelerators are essential for edge devices with limited resources and power constraints. While quantization and binarization can efficiently reduce hardware cost, they still rely on the conventional Artificial Neural Network (ANN) computation pattern. The recently proposed Kolmogorov-Arnold Network (KAN) presents a novel network paradigm built on learnable nonlinear functions. However, it is computationally expensive for hardware deployment. Inspired by KAN, we propose BiKA, a multiply-free architecture that replaces nonlinear functions with binary, learnable thresholds, introducing an extremely lightweight computational pattern that requires only comparators and accumulators. Our FPGA prototype on Ultra96-V2 shows that BiKA reduces hardware resource usage by 27.73% and 51.54% compared with binarized and quantized neural network systolic array accelerators, while maintaining competitive accuracy. BiKA provides a promising direction for hardware-friendly neural network design on edge devices.

Figures & tables

Fig. 1 : Difference between the computation of single neuron in MLP (up) and KAN (down)
DatasetModel TypeModel SizeFrequencyLUTFFDSPBRAM
WineMLP13,32,8,3 69749936172
KAN13,4,3 14684374741950132
Dry BeanMLP16,20,15,10,7 889411328170
KAN16,2,7 16775587345449111781
MashroomMLP8,64,64,2 1093218903174
KAN8,24,2 31122751337291162991347
TABLE I : Resource Consumption of MLP and KAN on FPGA Platform Based on Post-implementation in Previous Work
Fig. 2 : Difference between the ANN, BNN, KAN, and BiKA
Fig. 3 : Converting one learnable nonlinear function to a series of weighted learnable thresholds
Fig. 5 : Mixed all learnable thresholds with the sum of α\alpha
Fig. 7 : Computation in BiKA neurons when set quantization parameter, mm , as one.
Fig. 9 : Hardware difference of systolic array accelerators between BNN, QNN, and BiKA
DatasetMNIST [ deng2012mnist ]CIFAR-10 [ Krizhevsky2009 ]
NetworkTFCSFCLFCCNV
F64/F32/F10F256/F256/F256/F10F1024/F1024/F1024/F10C64/C64/P2/C128/C128/P2/C256/C256/P2/
F512/F512/F10
Q-KAN [ qkan ]98.52%---
KAN [ qkan ]98.02%---
BNN93.34%97.39%98.05%65.20%
TABLE II : Inference Accuracy of BNN, QNN, KAN, and BiKA on MLP and tiny CNN with MNIST Classification
More in this topic

Neural Network Hardware Acceleration

See all 38 papers

Explore similar work

CardsList

  1. GRAU: A Generic Runtime-Reconfigurable Activation-Requantization Unit for Quantized Neural Network Accelerators

    Feb 25, 2026Yuhao Liu, Salim Ullah, Akash KumarNeural Network Hardware AccelerationNeural Network Activation Functions

    Ruhr University Bochum, Germany · Dresden University of Technology, Germany · Center for Scalable Data Analytics and Artificial Intelligence (ScaDS.AI Dresden/Leipzig), Germany

  2. BiKAN: Restoring Collapsed Basis of Binary Kolmogorov--Arnold Networks

    Aug 2, 2026Kazi Ahmed Asif Fuad, Lizhong ChenBinary Neural NetworksKolmogorov-Arnold Networks

    Department of EECS Oregon State University Corvallis, OR 97331

  3. Physical Analogue Kolmogorov-Arnold Networks based on Reconfigurable Nonlinear-Processing Units

    Feb 7, 2026Manuel Escudero, Mohamadreza Zolfagharinejad, Sjoerd van den Belt +2Efficient Neural Network InferenceNeural Network Hardware Acceleration

    1NanoElectronics Group, MESA+ Institute and BRAINS Center for Brain-Inspired Computing, University of Twente, P.O. Box 217, Enschede,2026 7500 AE, The Netherlands. · 2CAESUniversity GroupofandTwente,BRAINSEnschede,Center 7500for Brain-InspiredAE, The Netherlands.Computing · Institute of Physics, University of Münster, Münster, 48149 Germany.