MLDS
收藏资源简介:
MLDS是一个由马里兰大学巴尔的摩分校的John Clemens创建的大型数据集,包含超过22万个经过训练的神经网络模型。这些模型通过MLC@Home项目,一个基于全球志愿者分布式计算的平台生成。数据集旨在通过直接分析神经网络的结构和权重,解决传统评估方法如损失函数难以检测的问题。MLDS数据集不仅用于模型间的比较,还用于研究模型与训练数据之间的关系,特别是在模型安全性和可追溯性方面的应用。
MLDS is a large-scale dataset created by John Clemens from the University of Maryland, Baltimore County, containing over 220,000 trained neural network models. These models are generated via the MLC@Home project, a distributed computing platform powered by global volunteers. This dataset is designed to address problems that traditional evaluation methods such as loss functions struggle to detect, through direct analysis of the structure and weights of neural networks. The MLDS dataset is used not only for comparing different neural network models, but also for researching the relationship between models and their training data, with specific applications in model security and traceability.




