Pretraining ECG Data with Adversarial Masking Improves Model Generalizability for Data-Scarce Tasks | Read Paper on Bytez