FeatureAssociation.md

Symbols and Functions > WolframModel and WolframModelEvolutionObject > Properties >

Feature Association

"FeatureAssociation" computes some features that can be handled by FeatureExtraction of the WolframModel evolution and returns an Association whose keys describe each feature computed. For now, it computes properties associated with these feature groups:

• "StructurePreservingFinalStateGraph": The Graph version of the "FinalState" as given by HypergraphToGraph using the "StructurePreserving" transformation
• "ObjectProperties": A list of properties of the WolframModel that return a numeric value. Right now the list of properties is: "CausalGraph", "EventsCount", "PartialGenerationsCount", "AllEventsDistinctElementsCount", "AllEventsEdgesCount", "CompleteGenerationsCount", "TerminationReason".

In[] := WolframModel[{{x, y}, {x, z}} -> {{x, z}, {x, w}, {y, w}, {z, w}}, {{0, 0}, {0, 0}}, 5]["FeatureAssociation"]

This property is useful for applying machine learning to Wolfram Models explorations:

In[] := BlockRandom[
  FeatureSpacePlot[#["FeatureAssociation"] -> Image[#["FinalStatePlot"], ImageSize -> Tiny] & /@
    (WolframModel[{{x, y}, {x, z}} -> {{x, z}, {x, w}, {y, w}, {z, w}}, #, 6] &) /@
      Table[RandomHypergraph[{3, 2}, 2], 10], LabelingFunction -> Callout], RandomSeeding -> 3
]

For Multiway Systems it only computes features associated with the causal graph, returning Missing["NotExistent", {"MultiwaySystem", "FinalState"}] for features related to "StructurePreservingFinalStateGraph", as there is no "FinalState" in a Multiway System:

In[] := WolframModel[{{x, y}, {x, z}} -> {{x, z}, {x, w}, {y, w}, {z, w}}, {{1, 1}, {1, 0}, {1, 1}}, 3,
  "EventSelectionFunction" -> "MultiwaySpacelike"]["FeatureAssociation"]