pyphi.cost.estimate_analysis#

pyphi.cost.estimate_analysis(substrate, state=None, *, subset=None, compute=None, limit=1000000, scope=None)[source]#

Count the workload of a single-system analysis, without running it.

Drives the same enumeration machinery pyphi.analyze() would use under the active configuration: the system partition scheme, the connectivity-pruned purview sets, and the mechanism partition scheme. No φ is computed and no state is needed — every counted quantity is state-independent.

Parameters:

  • substrate (Substrate) – The substrate to analyze.

  • state (optional) – Accepted so the call can mirror analyze(substrate, state). The counts do not depend on it; it is only checked to have one entry per unit.

  • subset (optional) – Node indices (or labels) of the candidate system; None uses the whole substrate.

  • compute (str or None, optional) – None estimates the full analysis; "sia" only the system-partition axis; "distinctions" only the distinction axis; "ces" the distinction axis plus the system-partition axis, since unfolding a cause-effect structure computes a system irreducibility analysis first.

  • limit (int, optional) – Work budget for the counting walk itself: purview evaluations and fresh partition enumerations each cost one unit, while memoized partition counts are free. A walk that exceeds the budget stops immediately and reports capped=True.

  • scope (CESScope, optional) – Restrict the counted mechanisms and purviews to the scope’s feasibility surface. Affects only the distinction axis; the system-partition count and the structural ceilings (possible_distinctions, possible_relations) are properties of the full system.

Returns:

AnalysisEstimate – The counted workload.

Raises:

ValueError – If compute is not "sia", "ces", "distinctions", or None, or if state does not have one entry per unit.

Return type:

AnalysisEstimate

Examples

>>> from pyphi import examples
>>> est = estimate_analysis(examples.basic_substrate())
>>> est.mechanisms
7
>>> est.system_partitions
22