QKAN: Quantum Kolmogorov-Arnold Networks

Abstract

The field of quantum machine learning persistently explores how learning models can take advantage of quantum implementations. Recently, a new neural network architecture, called Kolmogorov-Arnold Networks (KAN), has emerged, inspired by the Kolmogorov-Arnold theorem. We designed a quantum version of KAN (QKAN) by combining parameterized quantum circuits with powerful algorithmic subroutines.

In this talk, I will introduce classical KANs along with some established quantum linear algebra tools, which allow to take sums and products, and implement general functions on non-unitary matrices. With these tools in hand, we will construct the QKAN model and explore its potential applications.

References

• arxiv.org/abs/2410.04435