objectif

Vectorisation stateless par hashing (peu coûteuse mémoire).

code minimal

from sklearn.feature_extraction.text import HashingVectorizer

vec = HashingVectorizer(n_features=2**10, alternate_sign=False)
Xv = vec.transform(["a b c", "b c d"])
print(Xv.shape[1] == 1024)

utilisation

from sklearn.linear_model import SGDClassifier

clf = SGDClassifier().fit(Xv, [0,1])
print(hasattr(clf, "predict"))

variante(s) utile(s)

from sklearn.feature_extraction.text import HashingVectorizer
print(HashingVectorizer(n_features=2**8).transform(["x"]).shape[1] == 256)

notes

• Impossible d’inspecter les features (pas d’inverse mapping).