[bohyemianote: Oct.28.2019] IEEEVIS 2019

[All opinions are my own. This research was sponsored by the Microsoft AI Platform UX Design team.]

A few days ago, I took part in the IEEEVIS from Oct 20 to Oct 25 in Vancouver. Numerous studies in the context of visualization in a multitude of domains were introduced in the conference via Keynotes, Workshops, Tutorials, Paper presentations, Exhibitions, and Panel talks.

Among one of the hottest topics was undeniably “AI/ML + Visualization”. What appealed to me was that I was able to promptly run through a wide range of ongoing discourses in the AI/ML domain in a few days, covering “the role and method of visualization for the interpretability, explainability, and evaluation of ML systems”, “bias caused by ML systems”, “fairness of AI” and more. Throughout the process, I observed a significant amount of meaningful open questions such as “It appears that the closer collaboration among engineers, scientists, and designers is essential. If so, how do we define the role of designers in the veiled studies? Furthermore, what would be its synergistic effects? (personal curiosity)”, “how difference between the data analytics vs the interpretation/evaluation methods of ML systems?”, “the editability of neural network algorithms”, “Ways to evaluate human’s trust against a generated ML algorithms” and more. Meanwhile, a fresh keynote was “Zoom in: Features and Circuits as the Basic Unit of Neural Networks” presented by Chris Olah from Open AI. He argued that visualization techniques are beginning to reveal a rich world of interacting features inside neural networks in a similar way to the microscope opened up cellular biology that began characterizing and understanding specific proteins and genetic circuits. This intriguing point of view enabled me to step forward to the comprehension of a quintessential neural network principle by reminding me of my undergraduate studies in visual communication design and molecular biology.

Four studies were awarded as the best papers this year. Among them, the most favorable ones were “FlowSense : A Natural Language Interface for Visual Data Exploration within a Dataflow System (Authors: Bowen Yu, Claudio T. Silva)” and “Data Changes Everything: Challenges and Opportunities in Data Visualization Design Handoff (Authors: Jagoda Walny, Christian Frisson, Mieka West, Doris Kosminsky, Søren Knudsen, Sheelagh Carpendale, Wesley Willett)”. For one, “FlowSense…” is a natural language interface for dataflow visualization system that utilizes state-of-the-art natural language processing technique to assist the data flow diagram construction. Both the concept of the study and their demo was outstanding in representing neat interactions and graph interfaces. The second paper, “Data Changes…” proposed potential ideas for future tools for prototyping, testing, and communicating data-driven designs for the purpose of supporting more successful and data visualization design collaboration between designers and developers. As a whole, I consider myself quite an appropriate person who fully comprehends the gap between data characterization tools, visualization design tools, and development platforms given that I had experiences in both data-driven designs and traditional UX design in the field. Nonetheless, I will continuously adhere to programming languages such as Javascript and Python for data exploration, data wrangling, and prototyping due to the flexibility and effectiveness. I strongly believe that the concomitant of the novel tool is the exponential increase of the synergistic effect between designers and developers in the complex data visualization design projects for the future.

Besides, information retrieval and interactivity were discussed in the broad range of topics (particularly, texts and documents related papers) given that those domains practically include fundamental concepts of data analytics. Lastly, from the Arts Program where I assist as a program committee, a few favorable papers regarding AI and Environmental study such as “Mapping The Prelude: A Visualisation of Wordsworth’s Poetry (Author: Andrew Richardson)”, and “Dustmark and Ozone Tattoos: Autographic displays of air pollution (Author: Dietmar Offenhuber)” were archived for my personal research purposes.

In this article, I picked out studies/research work related to my experience and interests. The primary topics included “Visualization, AI, ML, Information retrieval, Network graph, Interactivity, Design, and Art “ If anyone has similar interests with me and aspires to find readings, the list may help them get inspiration.

If anyone requires the full list of papers and keynotes, please feel free to check the official websites.

2 Best papers

Authors: Bowen Yu, Claudio T. Silva

Video Preview | VIS 2019 Talk

[I] Data Changes Everything: Challenges and Opportunities in Data Visualization Design Handoff (J) (Best Paper Award)

Authors: Jagoda Walny, Christian Frisson, Mieka West, Doris Kosminsky, Søren Knudsen, Sheelagh Carpendale, Wesley Willett

Video Preview | VIS 2019 Talk

5 Keynotes

[VIS Capstone]

http://www.johannadrucker.net/

https://vimeo.com/369216256

[VDS]

https://www.youtube.com/watch?v=5w_rgBbwQHw

https://beenkim.github.io/

[Machine learning from User Interactions for Visualization and Analytics]

http://vialab.science.uoit.ca/

[Vis X AI: 2nd Workshop on Visualization for AI Explainability]

https://colah.github.io/

[Arts Program]

Website

38 Papers

[XAI and Fairness]

[V] Summit: Scaling Deep Learning Interpretability by Visualizing Activation and Attribution Summarizations (J)

Video Preview

https://fredhohman.com/summit/

[V] FairVis: Visual Analytics for Discovering Intersectional Bias in Machine Learning ©

Video Preview

[Interactive machine learning]

[V] Visual Interaction with Deep Learning Models through Collaborative Semantic Inference (J)

Video Preview

[Short Papers: VIS Meets Machine Learning]

TeleGam: Combining Visualization and Verbalization for Interpretable Machine Learning ©

Video Preview

Toward Perception-based Evaluation of Clustering Techniques for Visual Analytics ©

Video Preview

Visualizing RNN States with Predictive Semantic Encodings ©

Video Preview

FeatureExplorer: Interactive Feature Selection and Exploration of Regression Models for Hyperspectral Images ©

Video Preview

[VIS meet AI]

LassoNet: Deep Lasso-Selection of 3D Point Clouds (J)

Video Preview

[V] Facetto: Combining Unsupervised and Supervised Learning for Hierarchical Phenotype Analysis in Multi-Channel Image Data (J)

Authors: Robert Krüger, Johanna Beyer, Won-Dong Jang, Nam Wook Kim, Artem Sokolov, Peter Sorger, Hanspeter Pfister

Video Preview

[Large Data and Dimensionality Reduction]

[I] SolarView: Low Distortion Radial Embedding with a Focus (T)

Authors: Thom Castermans, Kevin Verbeek, Bettina Speckmann, Michel A. Westenberg, Rob Koopman, Shenghui Wang, Hein van den Berg, Arianna Betti

Video Preview

[Multivariate & Multidimensional Data]

[I] An Incremental Dimensionality Reduction Method for Visualizing Streaming Multidimensional Data (J)

Video Preview

[V] Selection Bias Tracking and Detailed Subset Comparison for High-Dimensional Data (J)

Video Preview

[Deconstruction]

[I] Critical Reflections of Visualization Authoring Systems (J)

Video Preview

https://vis-tools-reflections.github.io/

[I] Investigating Direct Manipulation of Graphical Encodings as a Method for User Interaction (J)

Video Preview

https://encodingstudy.github.io/

[Searching & Querying]

[I] Searching the Visual Style and Structure of D3 Visualizations (J)

Video Preview

[V] You can’t always sketch what you want: Understanding Sensemaking in Visual Query Systems (J)

Video Preview

http://zenvisage.github.io/

[Words & Documents]

[V] LDA Ensembles for Interactive Exploration and Categorization of Behaviors (T)

Video Preview

[V] Semantic Concept Spaces: Guided Topic Model Refinement using Word-Embedding Projections (J)

Video Preview

http://vialab.science.uoit.ca/portfolio/concept-spaces

[V] Topic-based Exploration and Embedded Visualizations for Research Idea Generation (T)

Video Preview

[Infographics & Storytelling]

[V] Multimodal Analysis of Video Collections: Visual Exploration of Presentation Techniques in TED Talks (T)

Video Preview

[V] Supporting Story Synthesis: Bridging the Gap between Visual Analytics and Storytelling (T)

Video Preview

http://simingchen.me/

http://geoanalytics.net/and/

[CG&A Session 2]

Mapping and Visualizing Deep-Learning Urban Beautification

Video Preview

http://goodcitylife.org/facelift/

[Short Papers: Perception, Cognition, and Visualization Design]

A Markov Model of Users’ Interactive Behavior in Scatterplots

Video Preview

Evaluating Ordering Strategies of Star Glyph Axes

Video Preview

Interactive Visualisation of Hierarchical Quantitative Data: an Evaluation

Video Preview

Thumbnails for Data Stories: Survey of Current Practice

Video Preview

[Planning and Situational Awareness]

[V] Interactive Learning for Identifying Relevant Tweets to Support Real-time Situational Awareness (J)

Video Preview

[Animation & Sports]

[I] A Comparative Evaluation of Animation and Small Multiples for Trend Visualization on Mobile Phones (J)

Video Preview

[I] A Comparison of Visualizations for Identifying Correlation Over Time and Space (J)

Video Preview

[Infovis Opening]

[I] Design by Immersion: A Transdisciplinary Approach to Problem-Driven Visualizations (J)

Video Preview

[I] Criteria for Rigor in Visualization Design Study (J)

Video Preview

[I] What is Interaction for Data Visualization? (J)

Video Preview

[I] Why Authors Don’t Visualize Uncertainty (J)

Video Preview

[Drawing Nodes and Edges]

[I] A Deep Generative Model for Graph Layout (J)

Video Preview

[I] Interactive Structure-aware Blending of Diverse Edge Bundling Visualizations (J)

Video Preview

[I] Persistent Homology Guided Force-Directed Graph Layouts (J)

Video Preview

[Arts program]

Mapping The Prelude: A Visualisation of Wordsworth’s Poetry

Video Preview

Infranet: A Geospatial Data-Driven Neuro-Evolutionary Artwork

Video Preview

Dustmark and Ozone Tattoos: Autographic displays of air pollution

Pictorial (PDF)

Video Preview

Hyemi Song, Sr. Designer (Data Vis)@Microsoft, Bohyemian Lab / Former Data Vis. Specialist@MIT Senseable City Lab, UX designer@Naver, MFA@RISD hyemisong.com