Supervised learning of the Jaynes-Cummings Hamiltonian.

Area of Science:

• Quantum mechanics
• Computational physics
• Machine learning

Background:

• Estimating Hamiltonian parameters is crucial for quantum system analysis.
• Deep neural networks (DNNs) offer potential for complex system modeling.

Purpose of the Study:

• To assess the effectiveness of DNNs in determining Jaynes-Cummings Hamiltonian parameters using only energy spectra.
• To evaluate model performance with both clean and noisy spectral data.

Main Methods:

• Utilized a vanilla DNN (vDNN) for noiseless energy spectra.
• Investigated the impact of input node count on vDNN accuracy.
• Employed a denoising U-Net in conjunction with vDNN to handle noisy spectra.

Main Results:

• vDNN error decreased with more input nodes in noiseless cases.
• vDNN showed limited resilience to Gaussian noise.
• The combined U-Net and vDNN model achieved up to a 77% error reduction on noisy data.

Conclusions:

• DNNs are effective for Hamiltonian parameter estimation from energy spectra.
• Integrating denoising networks enhances DNN robustness against noise in quantum spectral data.