When One Point Is Not Enough: Addressing Ambiguous Instances in Dimensionality Reduction by Splitting
Researchers have introduced a novel graph-based method to address ambiguous instances in dimensionality reduction, a common source of visual artifacts. This approach identifies data points that are highly similar to multiple, distinct neighborhoods in high-dimensional space. By replicating these ambiguous instances as multiple points in the projection, each placed within its relevant neighborhood, the method aims to more accurately represent the data's structure and reduce partial neighborhood embedding. AI
IMPACT Improves visualization of complex datasets by more accurately representing ambiguous data points.