增强的基于表示的采样,以有效地生成机器学习原子间潜力的数据集
Moritz R Schäfer1, Johannes Kästner1
1Institute for Theoretical Chemistry, University of Stuttgart, Pfaffenwaldring 55, Stuttgart 70569, Germany.
Journal of chemical theory and computation
|February 2, 2026
概括
增强的基于表示的抽样 (ERBS) 为机器学习的潜力产生了多样化的培训数据. 这种方法可以重建自由能量表面,并有效地改进自我扩散系数模拟.
科学领域:
- 计算化学计算化学
- 材料科学 材料科学 材料科学
- 机器学习 机器学习
背景情况:
- 机器学习的原子间潜力需要大量,多样化的数据集才能准确.
- 目前用于生成此类数据的方法可能在计算上昂贵且耗时.
研究的目的:
- 介绍基于增强表现的抽样 (ERBS),这是生成高质量的培训数据的新方法.
- 展示ERBS捕捉分子运动和重建自由能量表面的能力.
- 评估ERBS在生成液态水模拟数据方面的表现,并与传统方法进行比较.
主要方法:
- ERBS使用原子描述符的维度缩小来识别集体变量.
- 应用了一个偏差潜力,灵感来自飞行中的概率增强采样.
- 用高斯式时刻描述器来捕捉集体分子运动.
主要成果:
- ERBS成功地从短偏差轨迹高准确度重建自由能量表面.
- 用液态水的ERBS数据训练的模型产生与较大,传统数据集相比较的自我扩散系数.
- 与不确定性驱动动力学相比,ERBS显著提高了对配置空间的探索.
结论:
- 对于机器学习的原子间潜能,ERBS提供了一种高效有效的方法来生成结构多样化的训练数据.
- 该方法通过降低数据生成成本,加速了准确分子模型的开发.
- ERBS为推进分子模拟和材料发现提供了一个有前途的战略.
相关概念视频
Mechanical Efficiency of Real Machines
1.3K
The mechanical efficiency of a machine is a fundamental concept that describes how effectively a machine can convert input work into output work. According to this concept, the efficiency of a machine is equal to the ratio of the output work to the input work. An ideal machine, meaning a machine that has no energy losses, has an efficiency of one. This implies that the input work and the output work are equal.
However, in reality, no machine can be truly ideal, and all of them experience some...
However, in reality, no machine can be truly ideal, and all of them experience some...
1.3K
Design Example: Setting a Curve Using Design Data
236
Designing and plotting a curve using field data requires precise calculations and execution. A horizontal curve with a radius of 200 meters and an intersection angle of 20 degrees is established using the method of perpendicular offsets from the long chord. The long chord, which spans between the curve's endpoints, is calculated to be 69.46 meters in length. To maintain accuracy in plotting, intervals of 3 meters are selected along the chord.The engineer determines the offset distances for each...
236
State Space Representation
573
The frequency-domain technique, commonly used in analyzing and designing feedback control systems, is effective for linear, time-invariant systems. However, it falls short when dealing with nonlinear, time-varying, and multiple-input multiple-output systems. The time-domain or state-space approach addresses these limitations by utilizing state variables to construct simultaneous, first-order differential equations, known as state equations, for an nth-order system.
Consider an RLC circuit, a...
Consider an RLC circuit, a...
573
Control Volume and System Representations
1.6K
Two key frameworks are employed to analyze mass, energy, and momentum transfer: the control volume approach and the system approach. These frameworks offer different perspectives, depending on whether the focus is on a specific region in space (control volume approach) or a defined mass of fluid (system approach).
The control volume approach considers a stationary region in space through which fluid flows. This region is bounded by a control surface. For instance, in the case of water...
The control volume approach considers a stationary region in space through which fluid flows. This region is bounded by a control surface. For instance, in the case of water...
1.6K
Graphical Representation of Inequalities
208
The graph of the equation where y equals x squared forms a curve known as a parabola. This curve acts as a boundary in the coordinate plane, dividing it into distinct regions based on the relative position of points.When the equality sign in the equation is replaced with an inequality—such as greater than, less than, greater than or equal to, or less than or equal to—the graphical representation changes from a single curve into a broader shaded area that signifies the set of all...
208
Generation of Action Potential in Skeletal Muscles
8.8K
Every cell in the body maintains a membrane potential due to an uneven distribution of positive and negative charges across its plasma membrane. The membrane potential is measured in millivolts and quantifies the difference in charge across the membrane.
Like neurons, muscle cells are also regarded as excitable due to their capacity to change in response to stimuli, primarily due to voltage-gated ion channels embedded in their plasma membranes, which get activated by alterations in the...
Like neurons, muscle cells are also regarded as excitable due to their capacity to change in response to stimuli, primarily due to voltage-gated ion channels embedded in their plasma membranes, which get activated by alterations in the...
8.8K


