Rauber, Paulo E.Falcão, Alexandre X.Telea, Alexandru C.Enrico Bertini and Niklas Elmqvist and Thomas Wischgoll2016-06-092016-06-092016978-3-03868-014-7-https://doi.org/10.2312/eurovisshort.20161164https://diglib.eg.org:443/handle/10Many interesting processes can be represented as time-dependent datasets. We define a time-dependent dataset as a sequence of datasets captured at particular time steps. In such a sequence, each dataset is composed of observations (high-dimensional real vectors), and each observation has a corresponding observation across time steps. Dimensionality reduction provides a scalable alternative to create visualizations (projections) that enable insight into the structure of such datasets. However, applying dimensionality reduction independently for each dataset in a sequence may introduce unnecessary variability in the resulting sequence of projections, which makes tracking the evolution of the data significantly more challenging. We show that this issue affects t-SNE, a widely used dimensionality reduction technique. In this context, we propose dynamic t-SNE, an adaptation of t-SNE that introduces a controllable trade-off between temporal coherence and projection reliability. Our evaluation in two time-dependent datasets shows that dynamic t-SNE eliminates unnecessary temporal variability and encourages smooth changes between projections.Humancentered computingInformation visualizationComputing methodologiesDimensionality reduction and manifold learning10.2312/eurovisshort.2016116473-77