Data and code for 'CERA: A Framework for Improved Generalization of Machine Learning Models to Changed Climates'
收藏资源简介:
CERA_paper.tar.gz contains data for plots in the paper.CERA_trained_model.tar.gz contains trained model.CERA_code.zip contains code for the model. For figure 2,4,5,6: reweightFalseqv_Ftrest_baseline_bs1024_norh_noBMSE (baseline), reweightFalseqv_Ftrest_baseline_bs1024_rh_BMSE (RH+B baseline), Together_reweightFalse_sepa1_gt0.003warmup4000_std0_latent0.0001_CNN_0.003AE0.99_bs4096_noRH (CERA), Together_reweightFalse_sepa1_gt0.003warmup4000_std0_latent0_CNN_0.003AE0.99_bs4096_noRH (CERA_noAlign). For figure 7: run_MLP_72x180x48(baseline), run_CERA_72x180x48(CERA). For figure 8: run_MLP_mean_up_to_1800000_TABS_QV (baseline), run_CERA_mean_up_to_1800000_TABS_QV (CERA). For figure 3: janni_baseline_bs1024_noCI (baseline), janni_baseline_bs1024_CI (RH+B baseline), wy_together_sepa1_gt0.003warmup4000_janni_noCI_std0_latent0.001_CNN_0.003AEloss0.99_bs4096 (CERA)



