A recurrent Gaussian quantum network for online processing of quantum time series

Area of Science:

• Quantum Computing
• Machine Learning
• Quantum Communication

Background:

• Traditional machine learning struggles with quantum data due to measurement-induced state disturbance.
• Online processing of quantum temporal data is hindered by intermediate measurements.
• Existing methods for quantum output tasks often require fixed random parameters, limiting flexibility.

Purpose of the Study:

• To introduce and evaluate a Recurrent Gaussian Quantum Network (RGQN) for processing quantum time series.
• To demonstrate the RGQN's capability in enhancing quantum communication tasks.
• To address hardware restrictions and improve resource efficiency in quantum communication.

Main Methods:

• Developed a Recurrent Gaussian Quantum Network (RGQN) model.
• Trained all internal interactions of the RGQN, unlike previous reservoir computer models.
• Applied the RGQN to benchmark tasks and specific quantum communication challenges.
• Implemented a small-scale version of a task on Xanadu's Borealis photonic processor.

Main Results:

• The RGQN achieved higher performance on benchmark tasks compared to models with fixed parameters.
• The RGQN improved the transmission rate of quantum channels with memory effects.
• The RGQN effectively counteracted unwanted memory effects in quantum communication.
• Demonstrated resource efficiency and removal of hardware restrictions for quantum communication tasks.

Conclusions:

• The RGQN offers a flexible and high-performance approach for processing quantum time series data.
• RGQNs are effective in addressing key challenges in quantum communication, including enhancing transmission rates and mitigating memory effects.
• The developed model shows promise for practical applications in quantum communication and computing, with potential for real-world implementation on photonic processors.