clean_features.py

import numpy as np

from sklearn.base import BaseEstimator
from sklearn.feature_selection.base import SelectorMixin

class CleanFeatures(BaseEstimator, SelectorMixin):
    def __init__(self):
        self.sieve = []
    
    def _get_support_mask(self):
        assert len(self.sieve) > 0
        return self.sieve
    
    def fit(self, X, y=None):
        cur_sieve = [True] * X.shape[-1]
        
        for cur_index, this_col in enumerate(X.T):
            for that_col in X.T[:cur_index,:]:
                if np.sum(np.abs(this_col*that_col-1)) > 1e-8 : continue  
                cur_sieve[cur_index] = False
                break
        
        self.sieve = np.array(cur_sieve)
        self.sieve[0] = True
        return self