The Multivariate Embedding Report
The Multivariate Embedding report window contains a plot and one or more tables. The report enables you to visualize the low-dimensional space as well as review the algorithm parameters and computational details.
The form of the plot in the report depends on the number of dimensions specified in the Output Dimensions option in the launch window. If two dimensions are specified, the t-SNE Plot or UMAP Plot report appears. If three or more dimensions are specified, the 3D Plot report appears. These reports show a scatterplot in two or three dimensions of the first two or three embedded component scores for each observation, respectively.
The parameters table contains the parameter values that were specified in the launch window for either the UMAP method or the t-SNE method. See UMAP Options and t-SNE Options.
If the t-SNE method was specified in the launch window, the Iteration History table is also shown in the report. This table contains statistics from the computation iterations. Each row represents the values at the end of 100 successive iterations. The Iteration History table contains the following columns:
Iteration
The iteration number.
Cost
The value of the cost function at the corresponding iteration. This is the sum of the Kullback-Leibler divergences over all data points.
Max Gradient
The maximum value of the gradient over all data points at the corresponding iteration.
For more information about the columns in the Iteration History table, see Statistical Details for the Gradient Descent Algorithm.