File size: 2,366 Bytes
413dbdd | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 | """
precise_selectors.py
Custom sklearn transformers for the PRECISE-GBM pipeline.
Kept in a SEPARATE, IMPORTABLE module (NOT in the training script's __main__)
so that:
1. joblib / pickle can resolve the class by reference during parallel CV
(a class defined in __main__ raises
"Can't pickle <class '__main__.LassoSelector'>"), and
2. models saved with joblib.dump() can be RELOADED in a fresh session or in
retrain_helper.py, because the reference resolves to
precise_selectors.LassoSelector rather than __main__.LassoSelector.
Keep this file on the PYTHONPATH (simplest: same folder as the training and
retrain scripts).
"""
import numpy as np
from sklearn.base import BaseEstimator, TransformerMixin
from sklearn.linear_model import Lasso
class LassoSelector(BaseEstimator, TransformerMixin):
"""LASSO feature selection as a pipeline step (refit inside every CV fold).
Two-step alpha fallback: try each alpha in order, keep the first that
yields a non-empty support. If nothing survives any alpha for a fold,
keep ALL features and record it via `all_features_fallback_` so frequent
fallback (alphas too aggressive for the fold size) is auditable.
"""
def __init__(self, alphas=(0.1, 0.01), max_iter=10000, random_state=42):
self.alphas = alphas
self.max_iter = max_iter
self.random_state = random_state
def fit(self, X, y):
X = np.asarray(X)
y = np.asarray(y).ravel()
self.n_features_in_ = X.shape[1]
support = None
for alpha in self.alphas:
lasso = Lasso(alpha=alpha, max_iter=self.max_iter, random_state=self.random_state, tol=1e-4)
lasso.fit(X, y)
idx = np.flatnonzero(lasso.coef_ != 0)
if idx.size > 0:
support = idx
break
if support is None:
support = np.arange(self.n_features_in_)
self.all_features_fallback_ = True
else:
self.all_features_fallback_ = False
self.support_ = support
return self
def transform(self, X):
return np.asarray(X)[:, self.support_]
def get_support(self, indices=False):
if indices:
return self.support_
mask = np.zeros(self.n_features_in_, dtype=bool)
mask[self.support_] = True
return mask
|