Journal Article

FZJ-2026-02730

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png A Flexible and Energy‐Efficient Compute‐in‐Memory Accelerator for Kolmogorov–Arnold Networks

Sudarshan, C. (Corresponding author)FZJ* ; Artner, O.FZJ* ; Manea, P.-P.FZJ* ; Siegel, S.FZJ* ; Hoffmann-Eifert, S.FZJ* ; Dittmann, R.FZJ* ; Strachan, J. P. (Corresponding author)FZJ*

2026
Wiley-VCH Verlag GmbH & Co. KGaA Weinheim

This record in other databases:  

Please use a persistent id in citations: doi:10.1002/aisy.202501220  doi:10.34734/FZJ-2026-02730

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.

---

Contributing Institute(s):

1. Elektronische Materialien (PGI-7)
2. Neuromorphic Compute Nodes (PGI-14)
3. JARA-FIT (JARA-FIT)

Research Program(s):

1. 5233 - Memristive Materials and Devices (POF4-523) (POF4-523)
2. BMBF 16ME0398K - Verbundprojekt: Neuro-inspirierte Technologien der künstlichen Intelligenz für die Elektronik der Zukunft - NEUROTEC II - (BMBF-16ME0398K) (BMBF-16ME0398K)
3. DFG project G:(GEPRIS)528378584 - TRR 404: Zukunftsweisende Elektronik durch aktive Bauelemente in drei Dimensionen (Active-3D) (528378584) (528378584)

Appears in the scientific report 2026

---

Database coverage:
; ; ; ; Article Processing Charges ; Clarivate Analytics Master Journal List ; Current Contents - Engineering, Computing and Technology ; DEAL Wiley ; DOAJ Seal ; Essential Science Indicators ; Fees ; IF >= 5 ; JCR ; SCOPUS ; Science Citation Index Expanded ; Web of Science Core Collection

---