On linearity or non-linearity in machine learning for quantum chaotic dynamics

Key points

• This raises the question of whether machine learning can provide an effective alternative for predicting quantum dynamics.1
• We address this question by formulating quantum dynamics as a time-series forecasting problem, using a few-qubit PXP chain, realizable with Rydberg-atom arrays, as a benchmark.1
• We compare two contrasting architectures: an expressive nonlinear Transformer and DLinear, a simple linear forecasting model.1
• Remarkably, these results show that observables generated by complex quantum many-body dynamics can be forecast with high accuracy through a simple linear mapping from past to future observations.1

Sources (1)

• [1]On linearity or non-linearity in machine learning for quantum chaotic dynamics
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 06:00 PM

  Accurately simulating chaotic quantum many-body dynamics remains a major computational challenge for classical methods, due to the rapid buildup and spatial spreading of entanglement during the evolution.

  This raises the question of whether machine learning can provide an effective alternative for predicting quantum dynamics.

Extractive summary: sentences quoted from the sources.