Uniqueness Score

The uniqueness_score measures how distinctive a data point is relative to the rest of the dataset in the original high-dimensional embedding space. It is calculated by determining the average distance from each data point from its nearest neighbors (top 50 nearest). The average of these distances serves as the uniqueness_score, where a larger average distance indicates that the data point is farther from its neighbors and therefore more unique.

Note

To further assist users with data curation tasks, Kolena automatically calculated a number of metrics based on the embedding space details. Enable automatic embedding extractions or upload your own embeddings to utilize these scores.

Interpretation

The uniqueness_score identifies data points that are unique or underrepresented within your dataset. A higher uniqueness_score suggests that a data point resembles fewer points in the dataset, highlighting rare features. This can be particularly significant in outlier detection or help enhance the diversity of a dataset. By analyzing data points with high uniqueness_score, you can gain ensure that unique patterns are not overlooked, and make informed decisions regarding special cases.