arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.03729cs.LG
收藏

使用离策略对数色散正则化高效训练玻尔兹曼生成器

Efficient Training of Boltzmann Generators Using Off-Policy Log-Dispersion Regularization

Institute of Nanotechnology, Karlsruhe Institute of Technology, Kaiserstr. 12, 76131 Karlsruhe, Germany(纳米技术研究所,卡尔斯鲁厄理工大学,Kaiserstr. 12,76131 卡尔斯鲁厄,德国);Institute for Anthropomatics and Robotics, Karlsruhe Institute of Technology, Kaiserstr. 12, 76131 Karlsruhe, Germany(人机学与机器人研究所,卡尔斯鲁厄理工大学,Kaiserstr. 12,76131 卡尔斯鲁厄,德国)

Henrik Schopmans, Christopher von Klitzing, Pascal Friederich

AI总结:

提出离策略对数色散正则化(LDR),通过利用目标能量标签作为形状正则化器,结合标准数据训练目标,无需额外在策略样本,显著提升玻尔兹曼生成器的数据效率和最终性能,样本效率提升达一个数量级。

摘要:

从未归一化概率密度中采样是计算科学中的一个核心挑战。玻尔兹曼生成器是一种生成模型,能够在给定温度下从物理系统的玻尔兹曼分布中进行独立采样。然而,其实际成功依赖于数据高效的训练,因为模拟数据和目标能量评估成本都很高。为此,我们提出了离策略对数色散正则化(LDR),这是一种新颖的正则化框架,基于对数方差目标的推广。我们将LDR应用于离策略设置,与基于数据的标准训练目标相结合,无需额外的在策略样本。LDR通过利用目标能量标签形式的额外信息,充当能量景观的形状正则化器。所提出的正则化框架具有广泛的适用性,支持有偏或无偏的模拟数据集,以及在没有目标样本的情况下进行纯变分训练。在所有基准测试中,LDR都提高了最终性能和数据效率,样本效率提升高达一个数量级。

英文摘要:

Sampling from unnormalized probability densities is a central challenge in computational science. Boltzmann generators are generative models that enable independent sampling from the Boltzmann distribution of physical systems at a given temperature. However, their practical success depends on data-efficient training, as both simulation data and target energy evaluations are costly. To this end, we propose off-policy log-dispersion regularization (LDR), a novel regularization framework that builds on a generalization of the log-variance objective. We apply LDR in the off-policy setting in combination with standard data-based training objectives, without requiring additional on-policy samples. LDR acts as a shape regularizer of the energy landscape by leveraging additional information in the form of target energy labels. The proposed regularization framework is broadly applicable, supporting unbiased or biased simulation datasets as well as purely variational training without access to target samples. Across all benchmarks, LDR improves both final performance and data efficiency, with sample efficiency gains of up to one order of magnitude.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑