Application of Large Language Models to Quantum State Simulation

Shuangxiang Zhou,Ronghang Chen,Zheng An,Shi-Yao Hou

2024-10-30

Abstract:Quantum computers leverage the unique advantages of quantum mechanics to achieve acceleration over classical computers for certain problems. Currently, various quantum simulators provide powerful tools for researchers, but simulating quantum evolution with these simulators often incurs high time costs. Additionally, resource consumption grows exponentially as the number of quantum bits increases. To address this issue, our research aims to utilize Large Language Models (LLMs) to simulate quantum circuits. This paper details the process of constructing 1-qubit and 2-qubit quantum simulator models, extending to multiple qubits, and ultimately implementing a 3-qubit example. Our study demonstrates that LLMs can effectively learn and predict the evolution patterns among quantum bits, with minimal error compared to the theoretical output states. Even when dealing with quantum circuits comprising an exponential number of quantum gates, LLMs remain computationally efficient. Overall, our results highlight the potential of LLMs to predict the outputs of complex quantum dynamics, achieving speeds far surpassing those required to run the same process on a quantum computer. This finding provides new insights and tools for applying machine learning methods in the field of quantum computing.

Quantum Physics

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the high time and resource consumption in traditional quantum simulators when simulating quantum circuits. As the number of qubits increases, the resource consumption grows exponentially, which makes it very difficult and costly to use traditional methods to simulate large - scale quantum systems. To solve this problem, researchers have proposed an innovative method: using large - scale language models (LLMs) to simulate quantum circuits. This method aims to effectively process and retain the dependencies in long - time - series data through machine - learning techniques, especially the powerful feature - learning ability of LLMs, so as to achieve efficient prediction of quantum states.

Specifically, the paper describes in detail the process of constructing 1 - qubit and 2 - qubit quantum simulator models, and further extends to multiple qubits, and finally realizes the simulation of 3 - qubit quantum circuits. Research shows that LLMs can effectively learn and predict the evolution patterns between qubits, with minimal error compared to the theoretical output states. Even when dealing with quantum circuits containing an exponential number of quantum gates, LLMs still maintain computational efficiency. Overall, this research shows the potential of LLMs in predicting the output of complex quantum dynamics, and its speed is far faster than that required to run the same process on a quantum computer. This finding provides new insights and tools for applying machine - learning methods in the field of quantum computing.