Joshua A. Levine — Page 2

arXiv2019

NeuralCubes: Deep Representations for Visual Data Exploration

Visual exploration of large multidimensional datasets has seen tremendous progress in recent years, allowing users to express rich data queries that produce informative visual summaries, all in real time. Techniques based on data cubes are some of the most promising approaches. However, these techniques usually require a large memory footprint for large datasets. To tackle this problem, we present NeuralCubes: neural networks that predict results for aggregate queries, similar to data cubes. NeuralCubes learns a function that takes as input a given query, for instance, a geographic region and temporal interval, and outputs the result of the query. The learned function serves as a real-time, low-memory approximator for aggregation queries. NeuralCubes models are small enough to be sent to the client side (e.g. the web browser for a web-based application) for evaluation, enabling data exploration of large datasets without database/network connection. We demonstrate the effectiveness of NeuralCubes through extensive experiments on a variety of datasets and discuss how NeuralCubes opens up opportunities for new types of visualization and interaction.

cs.DB↗

arXiv2017

Visualizing Time-Varying Particle Flows with Diffusion Geometry

The tasks of identifying separation structures and clusters in flow data are fundamental to flow visualization. Significant work has been devoted to these tasks in flow represented by vector fields, but there are unique challenges in addressing these tasks for time-varying particle data. The unstructured nature of particle data, nonuniform and sparse sampling, and the inability to access arbitrary particles in space-time make it difficult to define separation and clustering for particle data. We observe that weaker notions of separation and clustering through continuous measures of these structures are meaningful when coupled with user exploration. We achieve this goal by defining a measure of particle similarity between pairs of particles. More specifically, separation occurs when spatially-localized particles are dissimilar, while clustering is characterized by sets of particles that are similar to one another. To be robust to imperfections in sampling we use diffusion geometry to compute particle similarity. Diffusion geometry is parameterized by a scale that allows a user to explore separation and clustering in a continuous manner. We illustrate the benefits of our technique on a variety of 2D and 3D flow datasets, from particles integrated in fluid simulations based on time-varying vector fields, to particle-based simulations in astrophysics.

cs.GR↗