Replication Data for: Interpreting learning dynamics of autoencoders: Transient scaling and emerging concepts of the Ising model

This dataset contains all relevant data to reproduce the results of the paper titled "Interpreting learning dynamics of autoencoders: Transient scaling and emerging concepts of the Ising model" and supplementary figures. "code.tar.xz" contains the code to generate the datasets and model checkpoints used in the analysis. Information on the code including instructions on how to reproduce the data can be found in the file README.md. "ising_data.tar.xz" contains the resulting datasets.

The datasets where generated with Markov-Chain-Monte-Carlo based on Glauber Dynamics of the Ising model and used for training and validation. "checkpoints.tar.xz" contains the resulting model checkpoints, including the used training configuration parameters and some metrics computed based on the datasets included in "ising_data.tar.xz". The model checkpoints where saved during training with different hyper-parameters. "figures.tar.xz" contains supplementary figures for the paper created based on the model checkpoints/metrics in "checkpoints.tar.xz" using the data from "ising_data.tar.xz".