LOCI fails on MacOS with Python 2.7 (caused by np.count_nonzero)

[yzhao062]

It is noted running LOCI model on MacOS with Python 2.7 may fail. One potential cause is the following code, as np.count_nonzero returns int instead of array.
I am currently investigating how to fix it. Please stay tuned.

 def _get_alpha_n(self, dist_matrix, indices, r):
        """Computes the alpha neighbourhood points.
        
        Parameters
        ----------
        dist_matrix : array-like, shape (n_samples, n_features)
            The distance matrix w.r.t. to the training samples.
        
        indices : int
            Subsetting index
        
        r : int
            Neighbourhood radius
            
        Returns
        -------
        alpha_n : array, shape (n_alpha, )
            Returns the alpha neighbourhood points.       
        """

        if type(indices) is int:
            alpha_n = np.count_nonzero(
                dist_matrix[indices, :] < (r * self._alpha))
            return alpha_n
        else:
            alpha_n = np.count_nonzero(
                dist_matrix[indices, :] < (r * self._alpha), axis=1)
            return alpha_n

The error message looks like below:

(test27) bash-3.2$ python loci_example.py
/anaconda2/envs/test27/lib/python2.7/site-packages/pyod/models/loci.py:199: RuntimeWarning: divide by zero encountered in double_scalars
outlier_scores[p_ix] = mdef/sigma_mdef
/Users/zhaoy9/.local/lib/python2.7/site-packages/numpy/core/_methods.py:101: RuntimeWarning: invalid value encountered in subtract
x = asanyarray(arr - arrmean)
On Training Data:
Traceback (most recent call last):
File "loci_example.py", line 133, in
evaluate_print(clf_name, y_train, y_train_scores)
File "/anaconda2/envs/test27/lib/python2.7/site-packages/pyod/utils/data.py", line 159, in evaluate_print
roc=np.round(roc_auc_score(y, y_pred), decimals=4),
File "/anaconda2/envs/test27/lib/python2.7/site-packages/sklearn/metrics/ranking.py", line 356, in roc_auc_score
sample_weight=sample_weight)
File "/anaconda2/envs/test27/lib/python2.7/site-packages/sklearn/metrics/base.py", line 77, in _average_binary_score
return binary_metric(y_true, y_score, sample_weight=sample_weight)
File "/anaconda2/envs/test27/lib/python2.7/site-packages/sklearn/metrics/ranking.py", line 328, in _binary_roc_auc_score
sample_weight=sample_weight)
File "/anaconda2/envs/test27/lib/python2.7/site-packages/sklearn/metrics/ranking.py", line 618, in roc_curve
y_true, y_score, pos_label=pos_label, sample_weight=sample_weight)
File "/anaconda2/envs/test27/lib/python2.7/site-packages/sklearn/metrics/ranking.py", line 403, in _binary_clf_curve
assert_all_finite(y_score)
File "/anaconda2/envs/test27/lib/python2.7/site-packages/sklearn/utils/validation.py", line 68, in assert_all_finite
_assert_all_finite(X.data if sp.issparse(X) else X, allow_nan)
File "/anaconda2/envs/test27/lib/python2.7/site-packages/sklearn/utils/validation.py", line 56, in _assert_all_finite
raise ValueError(msg_err.format(type_err, X.dtype))
ValueError: Input contains NaN, infinity or a value too large for dtype('float64').

[yzhao062]

The problem gets fixed by updating:

Python == 2.7.15
numpy == 1.15.1
sklearn == 0.19.2
scipy == 1.1.0

If you are running issues on Python 2 with Mac, try to update Python and dependent libraries.