Abstract: Computational notebooks have become a major medium for data exploration and insight communication in data science. Although expressive, dynamic, and flexible, in practice they are loose collections of scripts, charts, and tables that rarely tell a story or clearly represent the analysis process. This leads to a number of usability issues, particularly in the comprehension and exploration of notebooks. In this work, we design, implement, and evaluate Albireo, a visualization approach to summarize the structure of notebooks, with the goal of supporting more effective exploration and communication by displaying the dependencies and relationships between the cells of a notebook using a dynamic graph structure. We evaluate the system via a case study and expert interviews, with our results indicating that such a visualization is useful for an analyst’s self-reflection during exploratory programming, and also effective for communication of narratives and collaboration between analysts.

Abstract: The formation of social groups is defined by the interactions among the group members. Studying this group formation process can be useful in understanding the status of members, decision-making behaviors, spread of knowledge and diseases, and much more. A defining characteristic of these groups is the pecking order or hierarchy the members form which help groups work towards their goals. One area of social science deals with understanding the formation and maintenance of these hierarchies, and in our work, we provide social scientists with a visual analytics tool - PeckVis - to aid this process. While online social groups or social networks have been studied deeply and lead to a variety of analyses and visualization tools, the study of smaller groups in the field of social science lacks the support of suitable tools. Domain experts believe that visualizing their data can save them time as well as reveal findings they may have failed to observe. We worked alongside domain experts to build an interactive visual analytics system to investigate social hierarchies. Our system can discover patterns and relationships between the members of a group as well as compare different groups. The results are presented to the user in the form of an interactive visual analytics dashboard. We demonstrate that domain experts were able to effectively use our tool to analyze animal behavior data. While the primary application of PeckVis is to animal interaction data, we also demonstrate how other interaction-based data, such as a debate, can be effectively transformed and analyzed with our tool.

Abstract: Context Aware (CA) systems that model and adapt to the behaviors of their users have many real-world applications. CA systems generally require accurately labeled training data in order to learn parameters necessary to model users' context behavior. Unfortunately, it is challenging to gather sufficient realistic context data in a controlled study environment where labels can be reliably gathered. For this reason, recent studies have made use of in-the-wild context data, where data is gathered through some passive sensing device such as a smartphone while users are periodically required to supply their context labels. However context labels gathered through in-the-wild studies are not completely reliable as users are error prone and may provide incomplete or inaccurate labels for their data, which in turn makes it difficult to build robust CA models. We propose a visual analytics approach to finding instances of mislabelled data and cleaning human context data. Data visualizations enable analysts to highlight similar data to find patterns and anomalies in human behaviors. However, such an approach is challenging when working with erroneous human-labelled data as linking data based on similar context labels is flawed as the labels themselves are in question. To that end, we present DELFI, a visual analytics framework for identifying and relabeling mislabeled and unlabeled instances of user-labeled human context data. DELFI visually identifies likely-mislabeled instances by color-coding an anomaly score. Additionally, DELFI links instances based on a novel concept called Multi-Feature Similarity Linking, which helps to identify the likely true labels of mislabeled and unlabeled instances. We demonstrate the utility of our approach with detailed use cases and evaluation from domain experts.

Abstract: To optimize the performance and efficiency of HPC applications, programmers and analysts often need to collect various performance metrics for each computer at different time points as well as the communication data between the computers. This results in a complex dataset that consists of multivariate time-series and communication network data, thereby makes debugging and performance tuning of HPC applications challenging. Automated analytical methods based on statistical analysis and unsupervised learning are often insufficient to support such tasks without the background knowledge from the application programmers. To better explore and analyze a wide spectrum of HPC datasets, effective visual data analytics techniques are needed. In this paper, we present a visual analytics framework for analyzing HPC datasets produced by parallel discrete-event simulations (PDES). Our framework leverages automated time-series analysis methods and effective visualizations to analyze both multivariate time-series and communication network data. Through several case studies for analyzing the performance of PDES, we show that our visual analytics techniques and system can be effective in reasoning multiple performance metrics, temporal behaviors of the simulation, and the communication patterns.

Abstract: This paper presents an approach to analyzing orthogonal pairwise projections focusing on identifying observations that are significant in calculating the outliers of a scatterplot. We also propose a prototype, called Outliagnostics, to guide users when interactively exploring abnormalities in large time series. Instead of focusing on detecting outliers at each time point, we monitor and display the temporal discrepant signatures of each data entry concerning the overall distributions. Our prototype is designed to handle the types of doubly-multivariate data series in parallel. To highlight the benefits and performance of our approach, we illustrate and validate the use of Outliagnostics on real-world datasets of various sizes in different parallelism configurations.

Abstract: Semantic interaction is an technique relying upon the interactive semantic exploration of data. While manipulating data items within a visualization, an underlying model learns from the intent underlying these interactions and updates the parameters of the model controlling the visualization. In this work, we propose, implement, and evaluate a model which defines clusters within this data projection, then projects these clusters into a two-dimensional space using a “proximity∼similarity” metaphor. These clusters act as targets against which data values can be manipulated, providing explicit user-driven cluster membership assignments to train the underlying models. Using this cluster-first approach can improve the speed and efficiency of laying out a projection of high-dimensional data, with the tradeoff of distorting the global projection space.

Abstract: Rigorous data science is interdisciplinary at its core. In order to make sense of high-dimensional data, data scientists need to enter into a dialogue with domain experts. We present Glyphboard, a visualization tool that aims to support this dialogue. Glyphboard is a zoomable user interface that combines well-known methods such as dimensionality reduction and glyph-based visualizations in a novel, seamless, and immersive tool. While the dimensionality reduction affords a quick overview over the data, glyph-based visualizations are able to show the most relevant dimensions in the data set at one glance. We contribute an open-source prototype of Glyphboard, a general exchange format for high-dimensional data, and a case study with nine data scientists and domain experts from four exemplary domains in order to evaluate how the different visualization and interaction features of Glyphboard are used.

Abstract: We present Sherpa, a computational steering mechanism for progressive visual analytics that automatically prioritizes computations performed on the dataset based on the analyst's navigational behavior in the data. The intuition is that navigation in data space is an indication of the analyst's attention and thus interest in the data. Our example web-based client/server implementation of Sherpa provides computational modules for genomic data analysis, where independent computations calculate test statistics relevant to biological inferences about gene regulation in various tumor types and their corresponding normal tissues. The position and dimension of the navigation window on the genomic sequence over time is used to prioritize these computations to genomic regions favored by the user. In a study with experienced genomic and visualization analysts, we found that Sherpa provided comparable accuracy to the offline condition, where all computations were completed prior to analysis, while enabling shorter completion times. We also provide a second illustrative example on large-scale stock market analysis over time (Sherpa Stock).

Abstract: A chart sequence is used to describe a series of visualization charts generated in the exploratory analysis by data analysts. It provides not only information details in each chart but also a logical relationship among charts, which assists in interpreting the exploratory process, helping analysts understand the data and make the following decisions. We present a novel chart sequencing method based on reinforcement learning to capture the connections between charts in the context of analysis tasks. The proposed method formulates a chart sequencing procedure as an optimization problem, which seeks an optimal policy to sequencing charts for the specific analysis task. In our method, a novel reward function is introduced, which takes both the analysis task and the factor of human cognition into consideration. We evaluate our method under the application scenarios of visualization recommendation, sequencing charts for reasoning analysis results, and making a chart design choice.