This work presents a novel technique for real-time density estimation using adaptive bandwidth and GPUs. Density estimation and heatmaps are one of the most commonly used types of visualization for geo-referenced data; it allows the user to easily get insights from the data in a simple and straightforward way, by visualizing the density of a dataset. The most common way to calculate such density is through Kernel Density Estimation (KDE). KDE, however, has a severe shortcoming; it uses a constant smoothing parameter over the whole domain. Given a small smoothing parameter, low density areas may suffer from under-smoothing. If we choose a large smoothing parameter, high density areas may suffer from over-smoothing. To overcome this problem, we propose using an adaptive smoothing parameter that varies according to the density of points. Our approach is GPU-based, allowing interactive frame rates for streaming data without any pre-computation, even with datasets of over a million data points. We demonstrate the usefulness of our proposal through a set of synthetic and real datasets.


Moore Sloan Poster Research Project