Axiomatizing Techniques

[haz]

We'll want to formally represent all of the information we have on the extraction methods in MACQ. Both implemented and otherwise. This has many downstream benefits including (1) automatically determining gaps in the literature; (2) identifying techniques applicable for certain assumptions; etc.

A starting point collectively built during one of the MACQ meetings:

/*
T = (observer & ~slaf & ~amdn & ~arms & fo & ~noisy & XYZ) \/ (observer & ~slaf & ~amdn & ~arms & fo & ~noisy & ~XYZ) \/ ....

Q = (XYZ & ~noisy)
   T & Q | _extraction = observer \/ slaf
*/


{
    "token": "AtomicPartialObservation",
    "data-features": {
        "observability": "partial",
        "parameterized": "atomic (or 'objects' or 'objects-and-types')"
    },
}



{
    "technique": "slaf",
    "implemented": true,
    "issue": "https://github.com/QuMuLab/macq/issues/14",
    "description": "SLAF: Simultaneous Learning and Filtering",

    "primary-token": "observations.AtomicPartialObservation",

    "possible-tokens": {
        "implemented": ["observation.AtomicFullyObservableObservation"],
        "not-implemented": ["..."]
    },

    "extracted-model-features": {
        "determinism": "deterministic",
    },

    "generated-model-features": {
        "determinism": "non-deterministic",
    }
}