features.py 文件源码

def _calculate_sparse(self, X, y, categorical, metafeatures, helpers):
        import sklearn.decomposition
        rs = np.random.RandomState(42)
        indices = np.arange(X.shape[0])
        # This is expensive, but necessary with scikit-learn 0.15
        Xt = X.astype(np.float64)
        for i in range(10):
            try:
                rs.shuffle(indices)
                truncated_svd = sklearn.decomposition.TruncatedSVD(
                        n_components=X.shape[1] - 1, random_state=i,
                        algorithm="randomized")
                truncated_svd.fit(Xt[indices])
                return truncated_svd
            except LinAlgError as e:
                pass
        self.logger.warning("Failed to compute a Truncated SVD")
        return None


# Maybe define some more...