Effects of dissipation in reservoir computing using a spin qubit array

Shion Mifune,Taro Kanao,Tetsufumi Tanamoto

2024-10-10

Abstract:Reservoir computing (RC) is one of the hottest research topic as an application of many physical devices because the device characteristics can be used directly in computing sequences. Quantum RC is also a promising candidate for application in small-number qubit systems. Here, we propose a quantum RC based on the spin qubit system that reflects the status of the spin qubits in experiments comprising a one-dimensional qubit array. Spin qubits are coupled via the Heisenberg interaction, and data sequences are inputted to one of the spin qubits via pulsed rotations. By introducing dissipation, we obtained a relatively good performance in the quantum RC.

Quantum Physics,Disordered Systems and Neural Networks

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to explore the effect of introducing dissipation in Quantum Reservoir Computing (QRC) based on spin - qubit arrays. Specifically, the author proposes a method of using spin - qubit systems for quantum reservoir computing and improves the performance of quantum reservoir computing by introducing dissipation.

### Main research contents:
1. **Background introduction**:
- Quantum reservoir computing is a computational method that uses the natural dynamic characteristics of physical systems to process sequential data.
- Spin - qubits are an important platform for quantum computing, but there are challenges in practical applications due to the complexity of control and measurement.

2. **Problem description**:
- Traditional quantum reservoir computing methods usually need to accurately encode input data into the first qubit, which is difficult to achieve in actual spin - qubit systems.
- Another problem is that the CNOT gate operation in traditional spin - qubit systems is relatively expensive and impractical to use frequently.

3. **Solutions**:
- The author proposes a new method to control spin - qubits through pulse sequences instead of directly encoding input data.
- A dissipation mechanism is introduced to simulate the noise and attenuation processes in actual physical systems in order to improve the performance of quantum reservoir computing.

4. **Experimental design**:
- Experiments are carried out using one - dimensional and ring - arranged spin - qubit arrays.
- The performance of the system is evaluated through Short - Term Memory (STM) tasks and Nonlinear Autoregressive Moving Average (NARMA) tasks.

5. **Result analysis**:
- The experimental results show that the introduction of dissipation can significantly improve the performance of quantum reservoir computing, especially in NARMA tasks.
- The dissipation mechanism helps to reduce the long - term memory of the system, so as to better process new input data sequences.

### Conclusion:
By introducing the dissipation mechanism, the author has successfully improved the performance of quantum reservoir computing based on spin - qubit arrays. This method provides new ideas for the optimization of spin - qubits in future practical applications.