sNQS_rbm: Smooth Neural Quantum States for Real-Time Evolution

This repository provides the implementation of the s-NQS (smooth Neural Quantum States) method.

s-NQS introduces a continuous-time variational ansatz for real-time quantum dynamics using Chebyshev interpolation of neural network parameters. The method enables stable, global optimization of neural quantum states via Monte Carlo sampling.

Features

• Real-time evolution of quantum many-body systems via NQS
• Smooth parametrization using Chebyshev basis with global optimization
• Monte Carlo sampling with Metropolis-Hastings algorithm
• Optimizer: AdamW with PyTorch backend

Repository Structure

• model.py — Defines system Hamiltonians
• rbm.py — Restricted Boltzmann Machine architecture
• sNQS_rbm.py — s-NQS evolution with Chebyshev basis
• sampler.py — MCMC sampler implementation
• utils.py — Utility functions
• vmc.py — Variational Monte Carlo utilities
• test_*.py — Unit tests

Dependencies

• Python 3.12+
• NumPy
• PyTorch
• Matplotlib

Running the Code

To train an s-NQS model:

python main.py