Abstract:Despite rapidly growing interest in harnessing machine learning in the study of quantum many-body systems, training neural networks to identify quantum phases is a nontrivial challenge. The key challenge is in efficiently extracting essential information from the many-body Hamiltonian or wave function and turning the information into an image that can be fed into a neural network. When targeting topological phases, this task becomes particularly challenging as topological phases are defined in terms of non-local properties. Here we introduce quantum loop topography (QLT): a procedure of constructing a multi-dimensional image from the "sample" Hamiltonian or wave function by evaluating two-point operators that form loops at independent Monte Carlo steps. The loop configuration is guided by characteristic response for defining the phase, which is Hall conductivity for the cases at hand. Feeding QLT to a fully-connected neural network with a single hidden layer, we demonstrate that the architecture can be effectively trained to distinguish Chern insulator and fractional Chern insulator from trivial insulators with high fidelity. In addition to establishing the first case of obtaining a phase diagram with topological quantum phase transition with machine learning, the perspective of bridging traditional condensed matter theory with machine learning will be broadly valuable.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to use machine - learning techniques to identify topological phases in quantum many - body systems. Specifically, the main challenge faced by the authors lies in how to efficiently extract necessary information from the many - body Hamiltonian or wave function and transform it into an image form that can be input into a neural network. When the target is a topological phase, this task becomes more challenging because topological phases are defined by non - local properties. To overcome this problem, the authors introduced the Quantum Loop Topography (QLT) method, constructing multi - dimensional images in independent Monte Carlo steps by evaluating two - point operators that form loops. This method is particularly suitable for topological phases where characteristic responses such as Hall conductivity need to be detected. ### Key problems solved in the paper: 1. **Efficient information extraction**: Extract necessary information from the many - body Hamiltonian or wave function, especially for topological phases, where this information is non - local. 2. **Constructing images suitable for neural networks**: Convert the extracted information into an image form for input into a neural network for training and recognition. 3. **Identifying topological phases**: Use the QLT method to train a neural network to distinguish Chern insulators and fractional Chern insulators from trivial insulators with high precision. 4. **Drawing phase diagrams**: Use the trained neural network to quickly generate phase diagrams containing topological quantum phase transitions. ### Method overview: - **Quantum Loop Topography (QLT)**: Construct multi - dimensional images by evaluating two - point operators that form loops. The configurations of these loops are guided by the characteristic responses that define the phase, such as Hall conductivity. - **Neural network architecture**: Use a fully - connected neural network with a single hidden layer and train it to recognize different topological phases. - **Training data**: Use the Variational Monte Carlo (VMC) method to generate samples of the many - body ground state as training data. ### Main contributions: 1. **First successful application of machine learning**: In the study of topological quantum phase transitions, the machine - learning method was successfully applied for the first time to generate phase diagrams containing topological quantum phase transitions. 2. **Efficient identification of topological phases**: Even with a simple neural network architecture, different topological phases can be distinguished with high precision. 3. **High generality**: The QLT method is not only applicable to fermionic topological phases but can also be extended to other situations where non - local information is required, such as superconducting states and Z2 topological order. Through these methods, the authors demonstrated the great potential of machine learning in condensed - matter physics research, especially when dealing with complex topological phases.

Quantum Loop Topography for Machine Learning

Machine Learning Z2 Quantum Spin Liquids with Quasiparticle Statistics

Interpreting Machine Learning of Topological Quantum Phase Transitions

Learning Global Quantum Properties from Local Measurements with Neural Networks

Quaternion-based machine learning on topological quantum systems

Machine Learning Topological States

Machine Learning Topological Phases with a Solid-State Quantum Simulator.

Probing Transport in Quantum Many-Fermion Simulations Via Quantum Loop Topography

Learning topological defects formation with neural networks in a quantum phase transition

Deep learning of topological phase transitions from entanglement aspects: An unsupervised way

Topological quantum phase transitions retrieved through unsupervised machine learning

Machine Learning Topological Invariants with Neural Networks

Machine Learning of Quantum Phase Transitions

Topological Data Analysis Assisted Machine Learning for Polar Topological Structures in Oxide Superlattices

Machine Learning Topological Invariants of Non-Hermitian Systems

Unsupervised machine learning of topological phase transitions from experimental data

Unsupervised Learning of Non-Hermitian Photonic Bulk Topology

Visualizing quantum phases and identifying quantum phase transitions by nonlinear dimensional reduction

Deep learning of topological phase transitions from entanglement aspects

Investigating Quantum Many-Body Systems with Tensor Networks, Machine Learning and Quantum Computers