Tuan Nhon Dang

Stacking Graphic Elements to Avoid Over-Plotting

An ongoing challenge for information visualization is how to deal with over-plotting forced by ties or the relatively limited visual field of display devices. A popular solution is to represent local data density with area (bubble plots, treemaps), color(heatmaps), or aggregation (histograms, kernel densities, pixel displays). All of these methods have at least one of three deficiencies:1) magnitude judgments are biased because area and color have convex downward perceptual functions, 2) area, hue, and brightnesshave relatively restricted ranges of perceptual intensity compared to length representations, and/or 3) it is difficult to brush or link toindividual cases when viewing aggregations. In this paper, we introduce a new technique for visualizing and interacting with datasets that preserves density information by stacking overlapping cases. The overlapping data can be points or lines or other geometric elements, depending on the type of plot. We show real-dataset applications of this stacking paradigm and compare them to other techniques that deal with over-plotting in high-dimensional displays.

Visual pattern discovery using random projections

An essential element of exploratory data analysis is the use of revealing low-dimensional projections of high-dimensional data. Projection Pursuit has been an effective method for finding interesting low-dimensional projections of multidimensional spaces by optimizing a score function called a projection pursuit index. However, the technique is not scalable to high-dimensional spaces. Here, we introduce a novel method for discovering noteworthy views of high-dimensional data spaces by using binning and random projections. We define score functions, akin to projection pursuit indices, that characterize visual patterns of the low-dimensional projections that constitute feature subspaces. We also describe an analytic, multivariate visualization platform based on this algorithm that is scalable to extremely large problems.

ScagExplorer: Exploring Scatterplots by Their Scagnostics

A scatter plot displays a relation between a pair of variables. Given a set of v variables, there are v(v- 1)/2 pairs of variables, and thus the same number of possible pair wise scatter plots. Therefore for even small sets of variables, the number of scatter plots can be large. Scatter plot matrices (SPLOMs) can easily run out of pixels when presenting high-dimensional data. We introduce a theoretical method and a testbed for assessing whether our method can be used to guide interactive exploration of high-dimensional data. The method is based on nine characterizations of the 2D distributions of orthogonal pair wise projections on a set of points in multidimensional Euclidean space. Working directly with these characterizations, we can locate anomalies for further analysis or search for similar distributions in a large SPLOM with more than a hundred dimensions. Our testbed, ScagExplorer, is developed in order to evaluate the feasibility of handling huge collections of scatter plots.

ReactionFlow: an interactive visualization tool for causality analysis in biological pathways

BackgroundMolecular and systems biologists are tasked with the comprehension and analysis of incredibly complex networks of biochemical interactions, called pathways, that occur within a cell. Through interviews with domain experts, we identified four common tasks that require an understanding of the causality within pathways, that is, the downstream and upstream relationships between proteins and biochemical reactions, including: visualizing downstream consequences of perturbing a protein; finding the shortest path between two proteins; detecting feedback loops within the pathway; and identifying common downstream elements from two or more proteins.ResultsWe introduce ReactionFlow, a visual analytics application for pathway analysis that emphasizes the structural and causal relationships amongst proteins, complexes, and biochemical reactions within a given pathway. To support the identified causality analysis tasks, user interactions allow an analyst to filter, cluster, and select pathway components across linked views. Animation is used to highlight the flow of activity through a pathway.ConclusionsWe evaluated ReactionFlow by providing our application to two domain experts who have significant experience with biomolecular pathways, after which we conducted a series of in-depth interviews focused on each of the four causality analysis tasks. Their feedback leads us to believe that our techniques could be useful to researchers who must be able to understand and analyze the complex nature of biological pathways. ReactionFlow is available at https://github.com/CreativeCodingLab/ReactionFlow.

TimeSeer: Scagnostics for High-Dimensional Time Series

We introduce a method (Scagnostic time series) and an application (TimeSeer) for organizing multivariate time series and for guiding interactive exploration through high-dimensional data. The method is based on nine characterizations of the 2D distributions of orthogonal pairwise projections on a set of points in multidimensional euclidean space. These characterizations include measures, such as, density, skewness, shape, outliers, and texture. Working directly with these Scagnostic measures, we can locate anomalous or interesting subseries for further analysis. Our application is designed to handle the types of doubly multivariate data series that are often found in security, financial, social, and other sectors.

TimeArcs: Visualizing Fluctuations in Dynamic Networks

In this paper we introduce TimeArcs, a novel visualization technique for representing dynamic relationships between entities in a network. Force‐directed layouts provide a way to highlight related entities by positioning them near to each other Entities are brought closer to each other (forming clusters) by forces applied on nodes and connections between nodes. In many application domains, relationships between entities are not temporally stable, which means that cluster structures and cluster memberships also may vary across time. Our approach merges multiple force‐directed layouts at different time points into a single comprehensive visualization that provides a big picture overview of the most significant clusters within a user‐defined period of time. TimeArcs also supports a range of interactive features, such as allowing users to drill‐down in order to see details about a particular cluster. To highlight the benefits of this technique, we demonstrate its application to various datasets, including the IMDB co‐star network, a dataset showing conflicting evidences within biomedical literature of protein interactions, and collocated popular phrases obtained from political blogs.

PathwayMatrix: visualizing binary relationships between proteins in biological pathways

BackgroundMolecular activation pathways are inherently complex, and understanding relations across many biochemical reactions and reaction types is difficult. Visualizing and analyzing a pathway is a challenge due to the network size and the diversity of relations between proteins and molecules.ResultsIn this paper, we introduce PathwayMatrix, a visualization tool that presents the binary relations between proteins in the pathway via the use of an interactive adjacency matrix. We provide filtering, lensing, clustering, and brushing and linking capabilities in order to present relevant details about proteins within a pathway.ConclusionsWe evaluated PathwayMatrix by conducting a series of in-depth interviews with domain experts who provided positive feedback, leading us to believe that our visualization technique could be helpful for the larger community of researchers utilizing pathway visualizations. PathwayMatrix is freely available at https://github.com/CreativeCodingLab/PathwayMatrix.

TimeExplorer: Similarity Search Time Series by Their Signatures

The analysis of different time series is an important activity in many areas of science and engineering. In this paper, we introduce a new method feature extraction for time series and an application TimeExplorer for similarity-based time series querying. The method is based on eleven characterizations of line graphs presenting time series. These characterizations include measures, such as, means, standard deviations, differences, and periodicities. A similarity metric is then computed on these measures. Finally, we use the similarity metric to search for similar time series in the database.

FmFinder: Search and Filter Your Favorite Songs

Choices in music express our taste and personality. Different people have different collections of favorite songs. The explosive growth of digital media makes it easier to access any songs we want. Consequently, finding the songs best fit to our tastes becomes more challenging. Existing solutions record user patterns of listening to music, then make recommendation lists for users. By applying information visualization techniques to this problem, we are able to provide users with a novel way to explore their list of recommendations. Based on that knowledge, users can filter the songs according to their needs and compare the music tastes of different groups of people.

Using random projections to identify class-separating variables in high-dimensional spaces

Projection Pursuit has been an effective method for finding interesting low-dimensional (usually 2D) projections in multidimensional spaces. Unfortunately, projection pursuit is not scalable to high-dimensional spaces. We introduce a novel method for approximating the results of projection pursuit to find class-separating views by using random projections. We build an analytic visualization platform based on this algorithm that is scalable to extremely large problems. Then, we discuss its extension to the recognition of other noteworthy configurations in high-dimensional spaces.