The GMMFit function outputs a message and output_table. The message describes these properties:
| Property | Value |
|---|---|
| Output Table | Name of the output table to which the function outputs cluster information (output_table). |
| Algorithm Used | Algorithm that the function used—Basic GMM or DP-DMM. |
| Stopping Criterion | Why the function stopped—maximum iterations reached or convergence reached. |
| Delta Log Likelihood | Change in the mean log-likelihood for each data point between the next-to-last and the final iterations. |
| Number of Iterations | Number of iterations that the function performed before stopping. |
| Number of Clusters | Number of clusters in the GMM. |
| Covariance Type | Spherical, diagonal, tied, or full. |
| Number of Data Points | Number of data points in the data set. |
| Global Mean | Mean of the data set. |
| Global Covariance | Covariance of the data set. |
| Log Likelihood | Log-likelihood of the data, given the GMM. |
| Akaike Information Criterion | Akaike Information Criterion. |
| Bayesian Information Criterion | Bayesian Information Criterion. |
The output_table format depends on the PackOutput argument. For PackOutput('false'), the default, the following table describes output_table.
| Column Name | Data Type | Description |
|---|---|---|
| cluster_id | INTEGER | Identification number of the cluster |
| points_assigned | INTEGER | Number of points in the training data set assigned to the cluster |
| covariance_type | VARCHAR | Covariance type (specified by the CovarianceType argument) |
| weight | DOUBLE PRECISION | Weight assigned to the cluster |
| dim_n | DOUBLE PRECISION | Mean of the cluster n.
The table has one such column for each dimension (n has the values 1 through D). |
| cov_n | DOUBLE PRECISION | Covariance of the cluster n. Depends on the covariance type: 'spherical': Each row in a single covariance column contains a single DOUBLE PRECISION value. 'diagonal': There are D covariance rows, each containing a DOUBLE PRECISION value. 'tied' or 'full': There are D(D-1) covariance rows, each containing a DOUBLE PRECISION value. The table has one such column for each dimension (n has the values 1 through D). |
| determinant | DOUBLE PRECISION | Determinant of the covariance matrix |
| precision | VARCHAR | Precision matrix, the inverse of the covariance matrix. The precision matrix is serialized and stored to improve the performance of the function GMMPredict. |
For PackOutput('true'), the following table describes output_table.
| Column Name | Data Type | Description |
|---|---|---|
| cluster_id | INTEGER | Identification number of the cluster |
| points_assigned | INTEGER | Number of points in the training data set assigned to the cluster |
| covariance_type | VARCHAR | Covariance type (specified by the CovarianceType argument) |
| weight | DOUBLE PRECISION | Weight assigned to the cluster |
| mean | VARCHAR | A vector of D DOUBLE PRECISION values that specify the mean of each cluster (for example, [4.5, 2.3, 1.3)]. |
| covariance | VARCHAR | Depends on the covariance type: 'spherical': Each row in a single covariance column contains a single DOUBLE PRECISION value. 'diagonal': Each row contains a white-space separated list of D DOUBLE PRECISION values. 'tied' or 'full': Each row contains a white-space separated list of D*D DOUBLE PRECISION values. |
| determinant | DOUBLE PRECISION | Determinant of the covariance matrix |
| precision | VARCHAR | Precision matrix, the inverse of the covariance matrix. The precision matrix is serialized and stored to improve the performance of the function GMMPredict. |