Tutorials¶

Read the examples directly in the documentation. Each tutorial includes the code, saved results and figures, with the same navigation and theme as the rest of the site.

Tutorial	What it covers
Introduction	Prepare AnnData or arrays, align paired inputs, choose model modes and basic parameters
Original Gaussian experiment	Original NPZ files, frozen learned T, checkpoint replay and direct training
Synthetic data	Gaussian-bump comparisons and BranchSync on five paired branching snapshots
Splitting and growth	Branching trajectories and changing population mass
Input adapters	Arrays, dictionary, NPZ, CSV and AnnData inputs
Gastrulation	Train on paired data or load archived balanced and unbalanced 20k models into the original saved RNA UMAP
Held-out biological reconstruction	Optional W2 normalization, a held-out pancreas stage, 1,000 training iterations and shared UMAP reconstruction
Gastrulation dynamics and mass	Paired gastrulation data, a 1,000-step unbalanced model, E8.5 growth and flow with supplied cell types, and incoming/outgoing/reweighting decomposition for a chosen cell type and interval
Train and inspect a cross-modal map	Map-only gastrulation training, held-out pairs, shared ATAC UMAP and cell-type/stage consistency
scMultisim	Train T and balanced COATI
Lineage gene and peak dynamics	Later-start NMP cohorts, terminal lineage screening, observed gene/peak decoders, quality checks and molecular curves