Guided Stable Dynamic Projections

Projections aim to convey the relationships and similarity of high-dimensional data in a low-dimensional representation. Most such techniques are designed for static data. When used time-dependent data, they usually fail create stable suitable low dimensional We propose two dynamic projection methods (PCD-tSNE LD-tSNE) that use global guides steer points. This avoids unstable movement does not encode dynamics while keeping t-SNE's neighborhood preservation ability. PCD-tSNE scores good balance between stability, preservation, distance LD-tSNE allows creating customizable projections. compare our 11 other using quality metrics datasets provided by recent benchmark