TRACE: Tractable Routing Autoencoder for Clinical ECG
TRACE is proposed, a Tractable Routing Autoencoder for Clinical ECG, whose 32-dimensional clinical latent space is specified in advance from domain knowledge rather than discovered by optimization, and reconstructs the ECG through a decoder that permits latent perturbation.
Shun-Bo Jia, Run-Ze Ma, Haonan Lyu et al.
· 0 citations