Figure 2
From: Symplectic encoders for physics-constrained variational dynamics inference
Illustration of the DMM generative process. The latent random variable \(\textbf{z}_t\) a function of its previous instance \(\textbf{z}_{t-1}\) passed through the transmission model, enforcing a Markov property on the latent variable. At each time step, observations \(\textbf{x}_t\) are generated by the emission model.
