Mix-Key: graph mixup with key structures for molecular property prediction.

Molecular property prediction faces the challenge of limited labeled data as it necessitates a series of specialized experiments to annotate target molecules. Data augmentation techniques can effectively address the issue of data scarcity. In recent years, Mixup has achieved significant success in traditional domains such as image processing. However, its application in molecular property prediction is relatively limited due to the irregular, non-Euclidean nature of graphs and the fact that minor variations in molecular structures can lead to alterations in their properties. To address these challenges, we propose a novel data augmentation method called Mix-Key tailored for molecular property prediction. Mix-Key aims to capture crucial features of molecular graphs, focusing separately on the molecular scaffolds and functional groups. By generating isomers that are relatively invariant to the scaffolds or functional groups, we effectively preserve the core information of molecules. Additionally, to capture interactive information between the scaffolds and functional groups while ensuring correlation between the original and augmented graphs, we introduce molecular fingerprint similarity and node similarity. Through these steps, Mix-Key determines the mixup ratio between the original graph and two isomers, thus generating more informative augmented molecular graphs. We extensively validate our approach on molecular datasets of different scales with several Graph Neural Network architectures. The results demonstrate that Mix-Key consistently outperforms other data augmentation methods in enhancing molecular property prediction on several datasets.

Patterns of interest change in stack overflow.

Stack Overflow is currently the largest programming related question and answer community, containing multiple programming areas. The change of user's interest is the micro-representation of the intersection of macro-knowledge and has been widely studied in scientific fields, such as literature data sets. However, there is still very little research for the general public, such as the question and answer community. Therefore, we analyze the interest changes of 2,307,720 users in Stack Overflow in this work. Specifically, we classify the tag network in the community, vectorize the topic of questions to quantify the user's interest change patterns. Results show that the change pattern of user interest has the characteristic of a power-law distribution, which is different from the exponential distribution of scientists' interest change, but they are all affected by three features, heterogeneity, recency and proximity. Furthermore, the relationship between users' reputations and interest changes is negatively correlated, suggesting the importance of concentration, i.e., those who focus on specific areas are more likely to gain a higher reputation. In general, our work is a supplement to the public interest changes in science, and it can also help community managers better design recommendation algorithms and promote the healthy development of communities.