Neural-network decoders for measurement induced phase transitions

Area of Science:

• Quantum Information Science
• Condensed Matter Physics
• Machine Learning

Background:

• Open quantum systems exhibit exotic dynamical phases, including measurement-induced entanglement phase transitions.
• Probing these transitions typically requires infeasible experimental repetitions for large systems.

Purpose of the Study:

• To develop a machine learning-based method for detecting entanglement phase transitions in monitored quantum systems.
• To overcome the experimental limitations of traditional methods for studying these transitions.

Main Methods:

• Leveraging neural network decoders to determine the state of reference qubits based on measurement outcomes.
• Analyzing the purification dynamics of entangling reference qubits.
• Testing the approach on Clifford and Haar random circuits.

Main Results:

• Entanglement phase transitions manifest as significant changes in the learnability of the neural network decoder.
• The machine learning approach demonstrates complexity and scalability.
• The method successfully detects entanglement phase transitions.

Conclusions:

• Machine learning provides an efficient and scalable tool for detecting entanglement phase transitions in open quantum systems.
• This approach can be applied to generic experiments, overcoming previous experimental barriers.
• The study highlights the interplay between quantum dynamics and artificial intelligence.