斯塔根:一种基于生成性对抗网络的方法来提高分子图形生成的稳定性
Jinping Zou1, Jialin Yu1, Pengwei Hu1
1Department of Mathematics, School of Mathematics and Computer Sciences, Nanchang University, Nanchang, 330031, China; Institute of Mathematics and Interdisciplinary Sciences, Nanchang University, Nanchang, 330031, China.
这项研究介绍了STAGAN,一种新的深度生成模型,通过稳定训练来增强药物分子生成. STAGAN产生了更有效和独特的分子,解决了药物发现的深度学习中常见的问题.
科学领域:
- 计算化学是一种计算化学.
- 人工智能在药物发现中的作用
背景情况:
- 深度生成模型越来越多地用于分子生成的药物发现.
- 生成对抗性网络 (GAN) 可以学习分子结构,但会遭受训练不稳定,导致化学无效或单调的输出.
研究的目的:
- 提出一种新的方法,STAGAN,以解决训练分子生成深度生成模型的不稳定性问题.
- 提高药物发现中产生的分子的质量和多样性.
主要方法:
- 引入了STAGAN,将梯度惩罚术语纳入了歧视者.
- 在发电机中设计了一个并行批量规范化层,以提高训练稳定性.
主要成果:
- 与之前的模型相比,STAGAN证明了有效和独特分子的生成有所改善.
- 在QM9和ZINC-250K数据集上验证了性能.
结论:
- 拟议的STAGAN方法有效地解决了分子图形生成的深度生成模型中的训练不稳定性.
- STAGAN为药物发现分子图形生成的未来研究提供了改进的指导.
更多相关视频
08:21Curation of Computational Chemical Libraries Demonstrated with Alpha-Amino Acids
Published on: April 13, 2022
10:44Inherent Dynamics Visualizer, an Interactive Application for Evaluating and Visualizing Outputs from a Gene Regulatory Network Inference Pipeline
Published on: December 7, 2021
相关概念视频
Predicting Molecular Geometry
Relative Stabilities of Alkenes
Step-Growth Polymerization: Overview
Many natural and synthetic polymers are produced by...
Ziegler–Natta Chain-Growth Polymerization: Overview
Stability of Conjugated Dienes
A comparison of the enthalpies of hydrogenation of dienes reveals that conjugated dienes release less heat on hydrogenation, rendering them more stable than their nonconjugated analogs.
Stability of Substituted Cyclohexanes
The two chair conformations of cyclohexanes undergo rapid interconversion at room temperature. Both forms have identical energies and stabilities, each comprising equal amounts of the equilibrium mixture. Replacing a hydrogen atom with a functional group makes the two conformations energetically non-equivalent.
For example, in...
