Distribution fitting for the eigenvalues of the activations of the first layer.
收藏数据链接：
官方服务：
资源简介：
For all the four models, the eigenvalues are best parametrized by a Pareto distribution. We also find that the self-supervised models show smaller value for the shape parameter of the Pareto distribution.
创建时间：
2022-10-18
相关数据集
暂无相关数据集
