How to speed up featurization

[ppdebreuck]

A way to speed things up (but this might deteriorate the results) is to use a simpler featurizer. This is particularly useful for structures that are too slow to be featurized.

When creating a MODData you can specify a custom featurizer (e.g. basic_featurizer) :

md = MODData(materials=df["structure"].tolist(), structure_ids=df.index, featurizer=basic_featurizer)

You can make one by changing the Debreuck2020Featurizer

modnet/modnet/featurizers/presets/debreuck_2020.py

Line 9 in 6a5b139

class DeBreuck2020Featurizer(modnet.featurizers.MODFeaturizer):

For instance, make a BasicFeaturizer (here I removed the more computational features):

from modnet.featurizers.presets import DeBreuck2020Featurizer
class BasicFeaturizer(DeBreuck2020Featurizer):
    from pymatgen.analysis.local_env import VoronoiNN

    from matminer.featurizers.composition import (
        AtomicOrbitals,
        ElementFraction,
        ElementProperty,
        Stoichiometry,
        TMetalFraction,
        ValenceOrbital,
    )

    from matminer.featurizers.structure import (
        BondFractions,
        ChemicalOrdering,
        CoulombMatrix,
        DensityFeatures,
        EwaldEnergy,
        GlobalSymmetryFeatures,
        MaximumPackingEfficiency,
        RadialDistributionFunction,
        SineCoulombMatrix,
        StructuralHeterogeneity,
        XRDPowderPattern,
    )

    from matminer.featurizers.site import (
        AGNIFingerprints,
        AverageBondAngle,
        AverageBondLength,
        BondOrientationalParameter,
        ChemEnvSiteFingerprint,
        CoordinationNumber,
        CrystalNNFingerprint,
        GaussianSymmFunc,
        GeneralizedRadialDistributionFunction,
        LocalPropertyDifference,
        OPSiteFingerprint,
        VoronoiFingerprint,
    )

    oxid_composition_featurizers = ()

    composition_featurizers = (
        AtomicOrbitals(),
        ElementFraction(),
        ElementProperty.from_preset("magpie"),
        Stoichiometry(),
        TMetalFraction(),
        ValenceOrbital(),
    )

    structure_featurizers = (
        DensityFeatures(),
        GlobalSymmetryFeatures(),
        RadialDistributionFunction(),
        CoulombMatrix(),
        SineCoulombMatrix(),
        EwaldEnergy(),
        BondFractions(),
        StructuralHeterogeneity(),
        MaximumPackingEfficiency(),
        ChemicalOrdering(),
        XRDPowderPattern(),
    )

    site_featurizers = (
        AGNIFingerprints(),
        AverageBondAngle(VoronoiNN()),
        AverageBondLength(VoronoiNN()),
        BondOrientationalParameter(),
        ChemEnvSiteFingerprint.from_preset("simple"),
        CoordinationNumber(),
        CrystalNNFingerprint.from_preset("ops"),
        GaussianSymmFunc(),
        GeneralizedRadialDistributionFunction.from_preset("gaussian"),
        LocalPropertyDifference(),
        OPSiteFingerprint(),
        VoronoiFingerprint(),
    )

basic_featurizer = BasicFeaturizer()