Abstract:The advent of quantum computing holds the potential to revolutionize various fields by solving complex problems more efficiently than classical computers. Despite this promise, practical quantum advantage is hindered by current hardware limitations, notably the small number of qubits and high noise levels. In this study, we leverage adiabatic quantum computers to optimize Kolmogorov-Arnold Networks, a powerful neural network architecture for representing complex functions with minimal parameters. By modifying the network to use Bezier curves as the basis functions and formulating the optimization problem into a Quadratic Unconstrained Binary Optimization problem, we create a fixed-sized solution space, independent of the number of training samples. This strategy allows for the optimization of an entire neural network in a single training iteration in which, due to order of operations, a majority of the processing is done using a collapsed version of the training dataset. This inherently creates extremely fast training speeds, which are validated experimentally, compared to classical optimizers including Adam, Stochastic Gradient Descent, Adaptive Gradient, and simulated annealing. Additionally, we introduce a novel rapid retraining capability, enabling the network to be retrained with new data without reprocessing old samples, thus enhancing learning efficiency in dynamic environments. Experiments on retraining demonstrate a hundred times speed up using adiabatic quantum computing based optimization compared to that of the gradient descent based optimizers, with theoretical models allowing this speed up to be much larger! Our findings suggest that with further advancements in quantum hardware and algorithm optimization, quantum-optimized machine learning models could have broad applications across various domains, with initial focus on rapid retraining.

Towards Faster Reinforcement Learning of Quantum Circuit Optimization: Exponential Reward Functions

Cost Explosion for Efficient Reinforcement Learning Optimisation of Quantum Circuits

Reinforcement-Learning-Based Variational Quantum Circuits Optimization for Combinatorial Problems

Reinforcement Learning Based Quantum Circuit Optimization via ZX-Calculus

Quarl: A Learning-Based Quantum Circuit Optimizer

On the optimality of quantum circuit initial mapping using reinforcement learning

A Study on Optimization Techniques for Variational Quantum Circuits in Reinforcement Learning

Graph Neural Network Autoencoders for Efficient Quantum Circuit Optimisation

Optimization by a quantum reinforcement algorithm

Quantum-accessible reinforcement learning beyond strictly epochal environments

Optimizing Tensor Network Contraction Using Reinforcement Learning

Model-Free Quantum Control with Reinforcement Learning

Deep Reinforcement Learning with Reward Design for Quantum Control

Curriculum reinforcement learning for quantum architecture search under hardware errors

Optimizing Quantum Circuits, Fast and Slow

Optimizing Circuit Reusing and its Application in Randomized Benchmarking

Optimization of Reinforcement Learning Using Quantum Computation

Sparks of Quantum Advantage and Rapid Retraining in Machine Learning

Quantum-Inspired Reinforcement Learning for Quantum Control

Reinforcement Learning for Digital Quantum Simulation

Experimental quantum speed-up in reinforcement learning agents