pyphi.estimate.estimate_substrate#

pyphi.estimate.estimate_substrate(data, *, regime, prior=0.5, node_labels=None, model='counts')[source]#

Estimate a posterior over substrates from binary transition data.

Parameters:
  • data (ndarray[tuple[Any, ...], dtype[integer]] | tuple[ndarray[tuple[Any, ...], dtype[integer]], ndarray[tuple[Any, ...], dtype[integer]]]) – Either a pair (current, next) of integer arrays, each of shape (T, n) with one transition per row (the natural form for perturbational trials), or a single integer array of shape (T, n) — a trajectory whose transitions are the consecutive row pairs (the natural form for observational recordings). Both forms are accepted under both regimes; values must be 0/1.

  • regime (str) – Required assertion about how the data was produced: "perturbational" (current states set by intervention) or "observational" (passively recorded trajectory; see the module docstring for the identifiability caveat).

  • prior (float) – Symmetric Beta prior pseudocount added to every TPM cell (default Jeffreys, 1/2). Must be positive.

  • node_labels (Sequence[str] | None) – Optional labels for the substrate units.

  • model (str) – Estimation model; only "counts" is implemented.

Returns:

Return type:

SubstratePosterior