A Flexible and Energy‐Efficient Compute‐in‐Memory Accelerator for Kolmogorov–Arnold Networks
Open Access
- 1 March 2026
- journal article
- Published by Wiley in Advanced Intelligent Systems
- Vol. 8 (5)
Abstract
Emerging Kolmogorov–Arnold networks (KANs) replace the linear weights of neural networks with trainable nonlinear functions. This modification is particularly attractive for scientific computing, where KANs can match the accuracy of conventional multilayer perceptrons (MLPs) while reducing model size by up to 100×. However, this efficiency comes at the cost of computationally expensive nonlinear evaluations, unlike conventional MLPs dominated by linear matrix multiplications. We present a flexible and energy‐efficient compute‐in‐memory accelerator tailored for KANs, developed through cross‐layer optimization across algorithm, architecture, circuit, and device levels. The accelerator computes arbitrary nonlinear functions using a single‐read scheme and read‐optimized memory arrays with nonvolatile memristive devices. Our system achieves a lowest energy of 8.69 pJ per KAN function. In terms of energy‐delay product, it provides 1996× improvement over CPUs, 208× over standard MLP‐oriented compute‐in‐memory accelerators, and up to 71× over prior KAN accelerators. These results establish energy‐efficient hardware primitives for implementing advanced nonlinear networks in scientific computing.Keywords
Funding Information
- Deutsche Forschungsgemeinschaft (528378584)
- Bundesministerium für Forschung und Technologie (16ME0398K)
This publication has 46 references indexed in Scilit:
- DeepOKAN: Deep operator network based on Kolmogorov Arnold networks for mechanics problemsComputer Methods in Applied Mechanics and Engineering, 2025
- Fixing AI’s energy crisisNature, 2024
- Kinetic Monte Carlo simulation analysis of the conductance drift in Multilevel HfO2-based RRAM devicesNanoscale, 2024
- MemSPICE: Automated Simulation and Energy Estimation Framework for MAGIC-Based Logic-in-Memory2024
- Reliability Aspects of 28 nm BEOL‐Integrated Resistive Switching Random Access MemoryPhysica Status Solidi (a), 2023
- Nonvolatile Resistive Switching in Layered InSe via Electrochemical Cation DiffusionAdvanced Electronic Materials, 2021
- mMPU: Building a Memristor‐based General‐purpose In‐memory Computation Architecture2021
- Exploiting the switching dynamics of HfO2-based ReRAM devices for reliable analog memristive behaviorAPL Materials, 2019
- Approximate Fixed-Point Elementary Function Accelerator for the SpiNNaker-2 Neuromorphic Chip2018
- Taylor expansion of the accumulated rounding errorBIT Numerical Mathematics, 1976