Advanced Features

• Time-dependent parameters via extras_fn(t, y).
• Selective observation model: fit only some compartments or use different noise models per compartment.
• Checkpointing: save intermediate parameter states during long runs.
• Loss landscape visualization: grid-evaluate loss over parameter pairs and contour plot in log-scale.
• Posterior diagnostics: corner plots (e.g., using corner), trace plots, autocorrelation.