Factorized Structure Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/facebookresearch/pal/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集旨在评估在离散分解结构（FH）下多层感知机（MLPs）的性能，将模型表现与分解参数和超参数联系起来。数据生成时所涉及的参数包括输入因子、输出因子、父因子的数量、连通性参数以及集中度参数。该任务旨在评估MLPs在隐藏因子结构方面的性能表现。
This dataset is designed to evaluate the performance of Multi-Layer Perceptrons (MLPs) under the Discrete Factorization Hierarchy (FH), and correlate model performance with factorization parameters and hyperparameters. The parameters involved in data generation include input factors, output factors, the number of parent factors, connectivity parameters, and concentration parameters. This task aims to assess the performance of MLPs with respect to latent factor structures.
提供机构：
Authors of the paper
