seg1d.processing.Features.gen_weights

static Features.gen_weights(dataset)

Create weight table from reference data

Parameters:

dataset: List[Dict{feature:value}]

sets of references to be correlated against eachother to generate an importance score of each feature

Returns:

weightsDict{feature:score}

normalized sum of weights over all segments

See also

match_len

match length of all features

Notes

All features should be of same length.

Examples

>>> import numpy as np
>>> import seg1d.processing as process
>>> e10 = np.linspace(-np.pi*2, np.pi*2, 10)
>>> s1 = {'f1': np.sin(e10) , 'f2': np.sin(e10), 'f3': np.linspace(0,9,10) }
>>> s2 = {'f1': np.sin(e10) , 'f2': np.cos(e10), 'f3': -1*np.linspace(0,9,10) }
>>> s3 = {'f1': np.sin(e10) , 'f2': np.tan(e10), 'f3': np.ones((10))}
>>> d = [s1, s2, s3]
>>> r = process.Features.gen_weights(d)
>>> print([np.around(r[x], 3) for x in sorted(r)])
[array([1.]), array([-0.25]), array([-1.])]