Hierarchical and partition datasets
收藏NIAID Data Ecosystem2026-03-12 收录
下载链接：
https://zenodo.org/record/4898344
下载链接
链接失效反馈官方服务：
资源简介：
The datasets used throughout the paper "Learning Dynamics of Hierarchical Category Structure in Deep Non-Linear Networks", submitted to NeurIPS 2021.
These are synthetic datasets with hierarchical covariance structure (six hierarchies), P=936 features and C=64 categories (leaf node category). The hierarchical dataset is "synth_tree_nobj_64000_nclass_64_nhier_6_nfeat_936_beta_0.0001610_sigma_5.0000_norm_kemp_1_compressed.mat", while the partition dataset is "synth_tree_nobj_64000_nclass_64_nhier_6_nfeat_936_beta_0.0923671_sigma_5.0000_norm_kemp_1_compressed.mat".
The data were generated in a manner similar to Kemp & Tenenbaum, 2008 and Saxe et al., 2019.
创建时间：
2021-06-04
