聚焦科研团队特质的算法创新驱动力研究
Algorithm Innovation Driven by the Characteristics of Scientific Research Teams
目的/意义 将科研团队与算法创新结合，从科研团队特质的角度，探析其对算法创新的影响，有助于提升科研团队的学术能力，进而促进算法创新、科学创新。 方法/过程 建立科研团队规模、科研团队机构数量、科研团队机构类型三个科研团队测度指标以及算法性能与学术产出两个算法创新评价指标。以机器学习领域图像分类任务下的543个科研团队为例，运用非参数检验、多元线性回归模型探索科研团队特质对算法创新的影响效应，基于实证结果提出增强科研团队科研表现、促进算法再创新的建议。 结果/结论 科研团队特质测度指标均对算法创新存在影响效应，主要表现在：科研团队机构类型对算法创新具有显著影响效应，混合型科研团队在算法模型准确率上的表现最优，企业型科研团队在算法论文被引量上表现最优；科研团队规模对算法性能及学术产出均存在正向影响；科研团队机构数量对学术产出的影响呈正相关，对算法性能的影响呈负相关。<br />
计算技术、计算机技术
科研团队特质算法创新非参数检验多元线性回归
.聚焦科研团队特质的算法创新驱动力研究[EB/OL].(2023-04-14)[2025-11-05].https://chinaxiv.org/abs/202304.00978.点此复制
Purpose/Significance Combining scientific research team with algorithmic innovation and exploring its influence on algorithmic innovation from the perspective of scientific research team traits can help improve the academic capacity of scientific research team, which in turn can promote algorithmic innovation and scientific innovation. Method/Process Three research team measures, namely, research team size, number of research team institutions, and research team institution type, and two algorithm innovation evaluation indexes, namely, algorithm performance and academic output, are established. Taking 543 scientific research teams under the image classification task in the field of machine learning as an example, non-parametric tests and multiple linear regression models were used to explore the effect of scientific research team traits on algorithm innovation, and suggestions to enhance the scientific research performance of scientific research teams and promote algorithm reinvention were proposed based on the empirical results. Result/Conclusion All of the research team characteristics measures have an effect on algorithm innovation, mainly: the type of research team institution has a significant effect on algorithm innovation, the hybrid research team has the best performance in algorithm model accuracy, and the enterprise research team has the best performance in algorithm paper citations; the research team size has a positive effect on algorithm performance and academic output; the research team The number of institutions has a positive effect on the academic output and a negative effect on the algorithm performance.<br />
Research team characteristicsAlgorithm innovationNonparametric testMultiple linear regression
展开英文信息
评论