Neural Networks for Detecting Multimode Wigner Negativity.

Area of Science:

• Quantum Information Science
• Quantum Optics
• Machine Learning

Background:

• Characterizing quantum features in large Hilbert spaces is vital for quantum protocol testing.
• Continuous variable encoding, particularly quantum homodyne tomography, faces scalability issues due to exponentially increasing measurement requirements with the number of modes.

Purpose of the Study:

• To introduce a novel machine learning-based technique for directly detecting the negativity of the Wigner function in multimode quantum states.
• To overcome the intractability of conventional tomography methods for complex quantum systems.

Main Methods:

• Utilized a machine learning protocol employing artificial neural networks.
• Applied the method to numerically simulated multimode quantum states with known analytical Wigner functions.
• Tested the technique on an experimental multimode quantum state.

Main Results:

• The artificial neural network method accurately and rapidly detects Wigner function negativity.
• The approach demonstrates superior robustness compared to conventional methods, especially with limited data.
• The technique was successfully applied to an experimental quantum state, including resilience testing against losses.

Conclusions:

• The developed machine learning protocol offers an efficient and accurate solution for characterizing quantum features in multimode systems.
• This method significantly enhances the practical feasibility of testing quantum protocols involving complex quantum states.
• The technique shows promise for advancing quantum state characterization and quantum information processing.