search

byu-dml / metalearn

BYU's python library of useable tools for metalearning
MIT License
22 stars 6 forks source link

makes it easier to compare updated mfs #189

Closed emrysshevek closed 5 years ago

emrysshevek commented 5 years ago

Running compute_dataset_metafeatures.py now shows a breakdown of new, deleted, unchanged, and updated metafeatures to more easily compare when making changes to metafeatures.

Sample output (not necessarily real values or metafeatures :) )

38_sick_train_data.csv [(y)es, (v)erbose, (n)o]: v
UNCHANGED METAFEATURES
{
    "StdevStdDevOfStringLengthOfTextFeatures": {
        "compute_time": 4.8173824325203896e-05,
        "value": NaN
    },
    "kNN1NErrRate": {
        "compute_time": 0.2978278868831694,
        "value": 0.09093279580094993
    },
    "kNN1NKappa": {
        "compute_time": 0.2978278868831694,
        "value": 0.16702041294319114
    }
}
DELETED METAFEATURES
{
    "Useless Metafeaure": {
        "compute_time": 0.001676646017585881,
        "value": 52.35637003353562
    }
}
NEW METAFEATURES
{
    "Someone's new metafeature": {
        "compute_time": 1,
        "value": "Cool"
    }
}
UPDATED METAFEATURES
{
    "StdevCardinalityOfCategoricalFeatures": {
        "computed": {
            "compute_time": 0.0033262099605053663,
            "value": 123423
        },
        "known": {
            "compute_time": 0.0033224879880435765,
            "value": 0.7101612523427369
        }
    }
}
Runtime: 0.8915457725524902