AttentionViz: A Global View of Transformer Attention.

Transformer models are revolutionizing machine learning, but their inner workings remain mysterious. In this work, we present a new visualization technique designed to help researchers understand the self-attention mechanism in transformers that allows these models to learn rich, contextual relationships between elements of a sequence. The main idea behind our method is to visualize a joint embedding of the query and key vectors used by transformer models to compute attention. Unlike previous attention visualization techniques, our approach enables the analysis of global patterns across multiple input sequences. We create an interactive visualization tool, AttentionViz (demo: http://attentionviz.com), based on these joint query-key embeddings, and use it to study attention mechanisms in both language and vision transformers. We demonstrate the utility of our approach in improving model understanding and offering new insights about query-key interactions through several application scenarios and expert feedback.

Grand Challenges in Visual Analytics Applications.

In the past two decades, research in visual analytics (VA) applications has made tremendous progress, not just in terms of scientific contributions, but also in real-world impact across wide-ranging domains including bioinformatics, urban analytics, and explainable AI. Despite these success stories, questions on the rigor and value of VA application research have emerged as a grand challenge. This article outlines a research and development agenda for making VA application research more rigorous and impactful. We first analyze the characteristics of VA application research and explain how they cause the rigor and value problem. Next, we propose a research ecosystem for improving scientific value, and rigor and outline an agenda with 12 open challenges spanning four areas, including foundation, methodology, application, and community. We encourage discussions, debates, and innovative efforts toward more rigorous and impactful VA research.

Acquisition of chess knowledge in AlphaZero.

We analyze the knowledge acquired by AlphaZero, a neural network engine that learns chess solely by playing against itself yet becomes capable of outperforming human chess players. Although the system trains without access to human games or guidance, it appears to learn concepts analogous to those used by human chess players. We provide two lines of evidence. Linear probes applied to AlphaZero's internal state enable us to quantify when and where such concepts are represented in the network. We also describe a behavioral analysis of opening play, including qualitative commentary by a former world chess champion.

The What-If Tool: Interactive Probing of Machine Learning Models.

A key challenge in developing and deploying Machine Learning (ML) systems is understanding their performance across a wide range of inputs. To address this challenge, we created the What-If Tool, an open-source application that allows practitioners to probe, visualize, and analyze ML systems, with minimal coding. The What-If Tool lets practitioners test performance in hypothetical situations, analyze the importance of different data features, and visualize model behavior across multiple models and subsets of input data. It also lets practitioners measure systems according to multiple ML fairness metrics. We describe the design of the tool, and report on real-life usage at different organizations.

GAN Lab: Understanding Complex Deep Generative Models using Interactive Visual Experimentation.

Recent success in deep learning has generated immense interest among practitioners and students, inspiring many to learn about this new technology. While visual and interactive approaches have been successfully developed to help people more easily learn deep learning, most existing tools focus on simpler models. In this work, we present GAN Lab, the first interactive visualization tool designed for non-experts to learn and experiment with Generative Adversarial Networks (GANs), a popular class of complex deep learning models. With GAN Lab, users can interactively train generative models and visualize the dynamic training process's intermediate results. GAN Lab tightly integrates an model overview graph that summarizes GAN's structure, and a layered distributions view that helps users interpret the interplay between submodels. GAN Lab introduces new interactive experimentation features for learning complex deep learning models, such as step-by-step training at multiple levels of abstraction for understanding intricate training dynamics. Implemented using TensorFlow.js, GAN Lab is accessible to anyone via modern web browsers, without the need for installation or specialized hardware, overcoming a major practical challenge in deploying interactive tools for deep learning.

Deep learning of aftershock patterns following large earthquakes.

Aftershocks are a response to changes in stress generated by large earthquakes and represent the most common observations of the triggering of earthquakes. The maximum magnitude of aftershocks and their temporal decay are well described by empirical laws (such as Bath's law1 and Omori's law2), but explaining and forecasting the spatial distribution of aftershocks is more difficult. Coulomb failure stress change3 is perhaps the most widely used criterion to explain the spatial distributions of aftershocks4-8, but its applicability has been disputed9-11. Here we use a deep-learning approach to identify a static-stress-based criterion that forecasts aftershock locations without prior assumptions about fault orientation. We show that a neural network trained on more than 131,000 mainshock-aftershock pairs can predict the locations of aftershocks in an independent test dataset of more than 30,000 mainshock-aftershock pairs more accurately (area under curve of 0.849) than can classic Coulomb failure stress change (area under curve of 0.583). We find that the learned aftershock pattern is physically interpretable: the maximum change in shear stress, the von Mises yield criterion (a scaled version of the second invariant of the deviatoric stress-change tensor) and the sum of the absolute values of the independent components of the stress-change tensor each explain more than 98 per cent of the variance in the neural-network prediction. This machine-learning-driven insight provides improved forecasts of aftershock locations and identifies physical quantities that may control earthquake triggering during the most active part of the seismic cycle.

Visualizing Dataflow Graphs of Deep Learning Models in TensorFlow.

We present a design study of the TensorFlow Graph Visualizer, part of the TensorFlow machine intelligence platform. This tool helps users understand complex machine learning architectures by visualizing their underlying dataflow graphs. The tool works by applying a series of graph transformations that enable standard layout techniques to produce a legible interactive diagram. To declutter the graph, we decouple non-critical nodes from the layout. To provide an overview, we build a clustered graph using the hierarchical structure annotated in the source code. To support exploration of nested structure on demand, we perform edge bundling to enable stable and responsive cluster expansion. Finally, we detect and highlight repeated structures to emphasize a model's modular composition. To demonstrate the utility of the visualizer, we describe example usage scenarios and report user feedback. Overall, users find the visualizer useful for understanding, debugging, and sharing the structures of their models.

Visualizing Statistical Mix Effects and Simpson's Paradox.

We discuss how "mix effects" can surprise users of visualizations and potentially lead them to incorrect conclusions. This statistical issue (also known as "omitted variable bias" or, in extreme cases, as "Simpson's paradox") is widespread and can affect any visualization in which the quantity of interest is an aggregated value such as a weighted sum or average. Our first contribution is to document how mix effects can be a serious issue for visualizations, and we analyze how mix effects can cause problems in a variety of popular visualization techniques, from bar charts to treemaps. Our second contribution is a new technique, the "comet chart," that is meant to ameliorate some of these issues.

Participatory visualization with Wordle.

We discuss the design and usage of "Wordle," a web-based tool for visualizing text. Wordle creates tag-cloud-like displays that give careful attention to typography, color, and composition. We describe the algorithms used to balance various aesthetic criteria and create the distinctive Wordle layouts. We then present the results of a study of Wordle usage, based both on spontaneous behaviour observed in the wild, and on a large-scale survey of Wordle users. The results suggest that Wordles have become a kind of medium of expression, and that a "participatory culture" has arisen around them.

Mapping text with phrase nets.

We present a new technique, the phrase net, for generating visual overviews of unstructured text. A phrase net displays a graph whose nodes are words and whose edges indicate that two words are linked by a user-specified relation. These relations may be defined either at the syntactic or lexical level; different relations often produce very different perspectives on the same text. Taken together, these perspectives often provide an illuminating visual overview of the key concepts and relations in a document or set of documents.

The Word Tree, an interactive visual concordance.

We introduce the Word Tree, a new visualization and information-retrieval technique aimed at text documents. A word tree is a graphical version of the traditional "keyword-in-context" method, and enables rapid querying and exploration of bodies of text. In this paper we describe the design of the technique, along with some of the technical issues that arise in its implementation. In addition, we discuss the results of several months of public deployment of word trees on Many Eyes, which provides a window onto the ways in which users obtain value from the visualization.

Stacked graphs--geometry & aesthetics.

In February 2008, the New York Times published an unusual chart of box office revenues for 7500 movies over 21 years. The chart was based on a similar visualization, developed by the first author, that displayed trends in music listening. This paper describes the design decisions and algorithms behind these graphics, and discusses the reaction on the Web. We suggest that this type of complex layered graph is effective for displaying large data sets to a mass audience. We provide a mathematical analysis of how this layered graph relates to traditional stacked graphs and to techniques such as ThemeRiver, showing how each method is optimizing a different "energy function". Finally, we discuss techniques for coloring and ordering the layers of such graphs. Throughout the paper, we emphasize the interplay between considerations of aesthetics and legibility.

ManyEyes: a site for visualization at internet scale.

We describe the design and deployment of Many Eyes, a public web site where users may upload data, create interactive visualizations, and carry on discussions. The goal of the site is to support collaboration around visualizations at a large scale by fostering a social style of data analysis in which visualizations not only serve as a discovery tool for individuals but also as a medium to spur discussion among users. To support this goal, the site includes novel mechanisms for end-user creation of visualizations and asynchronous collaboration around those visualizations. In addition to describing these technologies, we provide a preliminary report on the activity of our users.

Designing for social data analysis.

The NameVoyager, a Web-based visualization of historical trends in baby naming, has proven remarkably popular. We describe design decisions behind the application and lessons learned in creating an application that makes do-it-yourself data mining popular. The prime lesson, it is hypothesized, is that an information visualization tool may be fruitfully viewed not as a tool but as part of an online social environment. In other words, to design a successful exploratory data analysis tool, one good strategy is to create a system that enables "social" data analysis. We end by discussing the design of an extension of the NameVoyager to a more complex data set, in which the principles of social data analysis played a guiding role.