Source

Gaussian MLE#

The Gaussian Maximum Likelihood Estimator (MLE) is able to spot outliers by computing a probability density function over the features assuming they are independent and normally (Gaussian) distributed. Assigning low probability density translates to a high anomaly score. The final anomaly score is given as the log likelihood of a sample being an outlier.

Interfaces: Estimator, Learner, Online, Ranking, Persistable

Data Type Compatibility: Continuous

Parameters#

# Param Default Type Description
1 threshold 5.0 float The minimum log likelihood to be flagged as an anomaly.
2 contamination 0.1 float The percentage of outliers that are assumed to be present in the training set.

Additional Methods#

Return the column means computed from the training set:

public means() : array

Return the column variances computed from the training set:

public variances() : array

Example#

use Rubix\ML\AnomalyDetection\GaussianMLE;

$estimator = new GaussianMLE(6.0, 0.1);

References#

  • T. F. Chan et al. (1979). Updating Formulae and a Pairwise Algorithm for Computing Sample Variances.