ALKPU: an active learning method for the DeePMD model with Kalman filter

TL;DR

This paper introduces ALKPU, an active learning method based on Kalman filter theory, to efficiently select training data for DeePMD neural network force fields, improving model accuracy and reducing computational costs.

Contribution

The paper proposes ALKPU, a novel active learning approach utilizing Kalman Prediction Uncertainty to enhance data selection for DeePMD models, leading to faster uncertainty reduction.

Findings

ALKPU effectively covers the configuration space in simulations.

It reduces training data requirements while maintaining accuracy.

The method improves training efficiency and computational resource usage.

Abstract

Neural network force field models such as DeePMD have enabled highly efficient large-scale molecular dynamics simulations with ab initio accuracy. However, building such models heavily depends on the training data obtained by costly electronic structure calculations, thereby it is crucial to carefully select and label the most representative configurations during model training to improve both extrapolation capability and training efficiency. To address this challenge, based on the Kalman filter theory we propose the Kalman Prediction Uncertainty (KPU) to quantify uncertainty of the model's prediction. With KPU we design the Active Learning by KPU (ALKPU) method, which can efficiently select representative configurations that should be labelled during model training. We prove that ALKPU locally leads to the fastest reduction of model's uncertainty, which reveals its rationality as a general active learning method. We test the ALKPU method using various physical system simulations and demonstrate that it can efficiently coverage the system's configuration space. Our work demonstrates the benefits of ALKPU as a novel active learning method, enhancing training efficiency and reducing computational resource demands.