转移一个分子基础模型用于聚合物性质预测
Pei Zhang1, Logan Kearney2, Debsindhu Bhowmik1
1Computational Sciences and Engineering Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States.
Journal of chemical information and modeling
|December 6, 2023
概括
大型语言模型 (LLM) 可以加速科学发现. 通过对小分子进行LLM预训练和对聚合物数据进行微调,研究人员可以在没有昂贵的数据增强的情况下实现准确的聚合物性质预测.
科学领域:
- 计算化学是一种计算化学.
- 材料科学是一种材料科学.
- 聚合物科学 聚合物科学
背景情况:
- 基于变压器的大型语言模型 (LLM) 在药物开发和材料发现等领域加速设计优化方面表现有前途.
- 自主监督的LLM预培训需要大量的数据集,这些数据在聚合物科学等专业领域很少.
- 目前用于聚合物的方法涉及数据增强,增加计算费用.
研究的目的:
- 通过LLMs研究转移学习对聚合物性质预测的有效性.
- 通过利用现有的小分子数据集来解决聚合物科学中的数据短缺问题.
- 将转移学习与传统数据增强技术的性能进行比较.
主要方法:
- 使用基于变压器的LLM,在广泛的小分子数据集上进行预训练.
- 在特定的聚合物特性数据集上微调预训练模型.
- 对聚合物进行基准预测任务的评估模型准确性.
- 结果与仅在增强聚合物数据上训练的模型进行比较.
主要成果:
- 在小分子上预训练和在聚合物上微调的LLM实现了与在增强聚合物数据上训练的模型相比的精度.
- 转移学习在克服聚合物科学中的数据短缺方面被证明是有效的.
- 这种方法为数据增强提供了一个计算效率高的替代方案.
结论:
- 使用LLM转移学习为聚合物性质预测提供了可行和高效的策略.
- 对小分子的预训练为在聚合物科学中应用的模型提供了坚实的基础.
- 这种方法减少了对大量数据增强的需求,节省了计算资源.
相关概念视频
01:01Polymers: Defining Molecular Weight
2.9K
Unlike small molecules with definite molecular weights, polymers are a mixture of individual polymer chains of varying lengths, each with a unique molecular weight. So, the molecular weight of a polymer is expressed as an average value based on the average size of the polymer chains. The two most common forms of averages used for polymers are the number average molecular weight and weight average molecular weight.
The number average molecular weight (Mn) is the summation of the number...
The number average molecular weight (Mn) is the summation of the number...
2.9K
01:10Polymers: Molecular Weight Distribution
3.4K
For any given polymer, the weight average molecular weight (Mw) is higher than, if not equal to, the number average molecular weight (Mn). The only situation in which the weight average molecular weight and the number average molecular weight are equal is when a polymer consists only of chains with equal molecular weight. However, this never happens in a synthetic polymer, since it is difficult to control the polymerization process up to a molecular level with accuracy to a hundred percent.
3.4K
01:08Molecular Weight of Step-Growth Polymers
2.2K
Step growth polymerization involves bi or multifunctional monomers. Bifunctional monomers react to form linear step growth polymers, whereas multifunctional monomers react to form non-linear or branched polymers.
As the step-growth polymerization involves step-wise condensation of monomers, the molecular weight also builds up eventually. Consequently, high molecular weight polymers are obtained at the late stages of the polymerization, where 99% of monomers have been consumed.
The extent of the...
As the step-growth polymerization involves step-wise condensation of monomers, the molecular weight also builds up eventually. Consequently, high molecular weight polymers are obtained at the late stages of the polymerization, where 99% of monomers have been consumed.
The extent of the...
2.2K
01:03Step-Growth Polymerization: Overview
3.5K
Step-growth or condensation polymerization is a stepwise reaction of bi or multifunctional monomers to form long-chain polymers. As all the monomers are reactive, most of the monomers are consumed at the early stages of the reaction to form small chains of reactive oligomers, which then combine to form long polymer chains in the late stages. Hence, the reaction has to proceed for a long time to achieve high molecular weight polymers.
Many natural and synthetic polymers are produced by...
Many natural and synthetic polymers are produced by...
3.5K
01:21Polymer Classification: Crystallinity
2.9K
Unlike ionic or small covalent molecules, polymers do not form crystalline solids due to the diffusion limitations of their long-chain structures. However, polymers contain microscopic crystalline domains separated by amorphous domains.
Crystalline domains are the regions where polymer chains are aligned in an orderly manner and held together in proximity by intermolecular forces. For example, chains in the crystalline domains of polyethylene and nylon are bound together by van der Waals...
Crystalline domains are the regions where polymer chains are aligned in an orderly manner and held together in proximity by intermolecular forces. For example, chains in the crystalline domains of polyethylene and nylon are bound together by van der Waals...
2.9K
02:00Molecular Models
38.5K
Physical models representing molecular architectures of chemical compounds play essential roles in understanding chemistry. The use of molecular models makes it easier to visualize the structures and shapes of atoms and molecules.
38.5K