Raw outputs for Empirical Study: Lerche et al. (2020)

The folders contain all the raw output of the empirical study from the paper:"Generalization in neural posterior estimation: Case studies with the racing diffusion model".The structure of the folders is as follows:- `experiment_*`: Study folder - `rdm_*`: Study condition folder (e.g., rdm_simple refers to the default settings) - `flow_matching`: Output related to neural posterior estimation (NPE) - `checkpoints`: Includes a `model.keras` file that contains the checkpoint of the NPE for this condition - `metrics`: Aggregated metrics based on posterior samples from NPE and MCMC - `npe_samples`: Raw posterior samples from NPE - `mcmc_samples`: Raw posterior samples obtained from Markov chain Monte Carlo (MCMC) - `test_data`: EmptyThe names of the study conditions are as follows:- `experiment_4`: - `rdm_simple`: Default settings with `mu_v_slope=1.5` and `beta_b=0.15` (condition medium-fixed) - `rdm_simple_lower`: Default settings with `mu_v_slope=2.5` and `beta_b=0.25` (condition lower-fixed) - `rdm_simple_meta`: Default settings with `mu_v_slope~U(0.5, 2.5)` and `beta_b~U(0.05, 0.25)` (condition full-varying) - `rdm_simple_lower`: Default settings with `mu_v_slope~U(0.5, 1.5)` and `beta_b~U(0.05, 0.15)` (condition lower-varying) - `rdm_simple_upper`: Default settings with `mu_v_slope~U(1.5, 2.5)` and `beta_b~U(0.15, 0.25)` (condition upper-varying) - `rdm_simple_upper`: Default settings with `mu_v_slope=0.5` and `beta_b=0.05` (condition upper-fixed)All datasets and posterior samples are stored in HDF5 files. Aggregated statistics are stored in CSV format.