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

2026-02-26

AI Summary

Edge devices, constrained by limited resources and power budgets, demand highly efficient and lightweight neural network acceleration solutions. Existing quantization or binarization approaches often rely on conventional computing paradigms, struggling to balance accuracy and efficiency. Inspired by Kolmogorov–Arnold Networks (KANs), this work proposes BiKA—an architecture that, for the first time, translates the KAN concept into a hardware-friendly design. BiKA eliminates multiplications entirely, replacing nonlinear functions with learnable binary thresholds and implementing a systolic-like dataflow using only comparators and accumulators. A prototype implemented on the Ultra96-V2 platform demonstrates significant hardware savings: 27.73% and 51.54% reductions in resource utilization compared to state-of-the-art binarized and quantized accelerators, respectively, while maintaining competitive accuracy.

Technology Category

• Machine Learning: Hardware-aware ML
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

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.

Problem

• lightweight neural network
• hardware accelerator
• edge devices
• Kolmogorov-Arnold Network
• binary neural network

Innovation

• Kolmogorov-Arnold Network
• multiply-free architecture
• binary learnable thresholds
• ultra lightweight accelerator
• edge AI hardware