Neural Network Matrix Product Operator: A Multi-Dimensionally Integrable Machine Learning Potential

Kentaro Hino,Yuki Kurashige

2024-11-01

Abstract:A neural network-based machine learning potential energy surface (PES) expressed in a matrix product operator (NN-MPO) is proposed. The MPO form enables efficient evaluation of high-dimensional integrals that arise in solving the time-dependent and time-independent Schrödinger equation and effectively overcomes the so-called curse of dimensionality. This starkly contrasts with other neural network-based machine learning PES methods, such as multi-layer perceptrons (MLPs), where evaluating high-dimensional integrals is not straightforward due to the fully connected topology in their backbone architecture. Nevertheless, the NN-MPO retains the high representational capacity of neural networks. NN-MPO can achieve spectroscopic accuracy with a test mean absolute error (MAE) of 3.03 cm$^{-1}$ for a fully coupled six-dimensional ab initio PES, using only 625 training points distributed across a 0 to 17,000 cm$^{-1}$ energy range. Our Python implementation is available at <a class="link-external link-https" href="https://github.com/KenHino/Pompon" rel="external noopener nofollow">this https URL</a>.

Machine Learning,Disordered Systems and Neural Networks,Statistical Mechanics,Chemical Physics,Quantum Physics

What problem does this paper attempt to address?

### What problem does this paper attempt to solve?

This paper aims to solve the difficult problem of high - dimensional integral calculation encountered when using machine learning (ML) methods to predict multi - dimensional potential energy surfaces (PES) in quantum chemistry and condensed - matter physics. Specifically, the authors proposed a neural - network - based matrix product operator (NN - MPO) model to efficiently evaluate high - dimensional integrals that appear when solving time - dependent and time - independent Schrödinger equations.

#### Main problems:

1. **Efficiency problem of high - dimensional integral calculation**:
- In many - body quantum simulations, calculating multi - dimensional integrals between wave functions and Hamiltonians is one of the bottlenecks. Traditional neural networks such as multi - layer perceptrons (MLP), due to their fully - connected structures, are inefficient in handling high - dimensional integrals and usually need to rely on approximate methods such as Monte Carlo integration.

2. **Overcoming the curse of dimensionality**:
- The computational complexity of high - dimensional integrals increases exponentially with the increase in dimension, which is the so - called "curse of dimensionality". Traditional MLP methods are difficult to effectively deal with this problem.

3. **Maintaining high - expressive ability**:
- Although MLP has strong expressive ability, it faces challenges when handling high - dimensional integrals. Therefore, how to improve the computational efficiency of high - dimensional integrals while maintaining high - expressive ability is an important problem.

#### Solutions:

The NN - MPO model proposed by the authors combines the advantages of tensor networks (TN) and can solve problems in the following aspects:

- **Efficiently evaluating multi - dimensional integrals**: By introducing the matrix product operator (MPO) form, NN - MPO can efficiently evaluate multi - dimensional integrals, thereby significantly improving computational efficiency.

- **Maintaining high - expressive ability**: NN - MPO retains the high - expressive ability of neural networks and can accurately fit complex potential energy surfaces with a small number of parameters.

- **Applicable to large - scale systems**: This model can be applied to systems with more than six dimensions, and high - precision prediction can be achieved with only a small number of training data points. For example, for a six - dimensional ab initio PES, a spectral - level accuracy (MAE of 3.03 cm⁻¹) can be achieved using 625 training points.

In summary, this paper successfully solves the problems of low efficiency in high - dimensional integral calculation and the curse of dimensionality by introducing the NN - MPO model, while maintaining high - expressive ability, providing new tools and methods for many - body quantum simulations.