Archive for October 2008
As part of Visweek, there were a few Panels which I have consistently enjoyed through all these years. The first panel was on
Grand Challenges for Information Visualization
Tamara Munzner gave an interesting talk which resonated with that I believe is necessary for our field. She emphasized on the fact that we “need open software for open data.” On numerous occasions, we see wonderful visualization techniques at conferences and in the IEEE TVCG journal, but rarely do we see the source code with some sample datasets being made available. I particularly applaud efforts such as Many eyes, VisTrails and some other visualization toolkits (which I cant think of right now) that allow not only visualization but also some information regarding how the visualization was created. Vistrails goes much farther and even provides detailed information regarding the steps taken by a user to reach a certain point.
Tamara also proposed a common framework which seems inspired by the field of security and software engineering. It was an interesting way to deal with providing visualization solutions to real world application domain problems. Here slides can be found at www.cs.ubc.ca/~tmm/talks/vis08/vis08.pdf
I think that some of the ideas that they discussed in the panel are crucial to our field of visualization (not just infovis). We need to make sure that we have something more than a long list of papers (textbooks/introductory 1-day workshops at conferences for first time attendees) for new students and practitioners. I am glad to see such enthusiasm and fervor as was palpable at the panel and hope to see many more such events that make us think as a community. I wasnt able to attend the entire panel and so if any of you have any comments on the same, please feel free to add them here.
Building a Research Group in Visualization
Panelists: Hamish Carr, Sheelagh Carpendale, Thomas Ertl, Helwig Hauser, Chris Johnson, Min Chen, Stephen North
This panel was hosted by Hamish Carr, who is at the University College Dublin. The panel started out with each of the esteemed panelists discussing what worked and how things worked out for them as a researcher as well as a research group.
Sheelagh Carpendale spoke first and basically said that she had identified five components to ‘success’ as regards forming a creative, productive research group. The five components are (i) Collaboration – Where multiple students collaborate and get more done by helping each other out. (ii) Competition – where students or sub-groups within the lab compete in a healthy manner towards evaluating techniques, developing software modules etc. (iii) Mentoring – Each new student is paired up with a senior member in the lab to help with adjusting to the lab as well as getting up to speed with research in the field. (iv) Individualism – It is critical to identify individual pieces for students so that they can claim ownership of a part of the project. She mentioned that its also important to encourage students to think on their own. (v) Scaffolding – Last but not the least, scaffolding is the glue that provides a productive environment to the students and researchers in the lab. Providing students with sufficient, high quality resources helps them achieve their goals as well as helps the faculty member achieve their goals.
Min Chen focused on challenges of leading a smallish research group. He spoke about how the Research Assessment Exercise (RAE) process helped and shaped his research group as well as the department’s growth at Swansea University, UK. Particularly, he spoke about how he read books on managing groups such as ‘Handbook of Small Group Research‘ by A. Paul Hare, which apparently is out of circulation. The amusing part of his talk was that the book had examined criminal gangs and figured out some of the rules and ideas about what works and doesnt work from observing gang leaders and the gang as a whole. He also spoke on how their group outings are mostly research focused and that has helped them grow as well as form connections with other faculty in the department.
Thomas Ertl spoke next on how his experience in astrophysics as well as industry helped him immensely. He said that having started his own company before becoming a faculty member, convinced him of the need of selling. By selling he meant selling an idea, a concept and so on. He said that writing skills can definitely be improved and there is no excuse to poor communication skills. One needs to be able to convince the reader/person sitting in front of you of the viability of your idea/algorithm/system. He said that “Success is a combination of individual performance and how others perceive you.” He said that in his lab he has always encouraged collaboration as opposed to competition.
Helwig Hauser spoke about the balance between demands (financial, research, departmental) as a professor versus own choices as regards research, research topic etc. He said that one should always keep an eye on the practical assets when conducting research. He defined them as software, algorithms and tools. He also spoke about the need to balance reactive vs proactive approaches to research. In some cases, one needs to work on a project to fund the research that one is more interested in.
Chris Johnson discussed how his small research lab of one student grew to a small research group and that grew into a large research group which in turn grew into a centre and now its a huge institute (Scientific Computing and Imaging Institute). He said that only through collaborations and having a big group of smart people around you, can you do more interesting science than just by working by yourself. He spoke abut how its always challenging to manage large groups but he said that they are very careful when hiring new faculty/researchers since one disgruntled person can make the environment unproductive.
Stephen North from AT&T Research labs gave the industry perspective to managing research groups. He said that the funding is more or less stable but its very important to have executive support from within the company. He said that long term goals and views are important for a research group, but also said that its not always possible to meet all the long term goals.
This was followed by an excellent discussion where attendees asked panelists some insightful questions. Instead of trying to summarize the interaction, I would like to direct you to Carlos Scheidegger’s excellent summarization of the Q&A session that followed.
Unfortunately, I could not make it to the Visual Analytics panel. Please let me know if any of you attended that session.
What Art Can Tell Us About the Brain
Margaret S. Livingstone, Professor of Neurobiology, Harvard Medical School
I knew the keynote talk at the IEEE Visualization conference was going to be interesting when the student volunteer at the door handed me stereo glasses before the talk. The talk was focused on introducing concepts about color and luminance that artists have been using effectively for hundreds of years.
In the first part of the talk, she focused on introducing concepts such as centre-surround.
Some of the highlights from her talk are as follows:
- In some cases, color contrast is not equal to luminance contrast.
- People are good at recognizing objects from different points of view. This might be interesting since there seems to be considerable amount of work in graphics and visualization on finding the ‘best’ view for a particular dataset.
- Equiluminance – In the paper “Vernier and Displacement Thresholds in Equiluminance” by Masami Funakawa, equiluminance is introduced as follows
The notion of equiluminance is based upon the assumption that in the human visual system there are two kinds of visual pathway, i.e., luminance and chromatic pathway. An equiluminous stimulus varies in color, but not in luminance, so that it is assumed to be signaled by the chromatic system, but not by the luminance system. In psychophysical studies these stimuli are used to isolate and probe the chromatic system.
This webpage describes the use of equiluminance in art and how luminance affects our perception http://www.webexhibits.org/colorart/anuszkiewicz.html
- Depth can be conveyed using motion, shading, perspective projection, occlusion and stereopsis.
- Luminance contrast: As long as the luminance is appropriate, shape can be conveyed. She showed some interesting examples of paintings [Matisse's The Woman in a Hat] and visual representations, where the colors were unnatural but it did not affect the perceived shape of face/object.
- Yellow, Blue and White, when used appropriately can convey motion in static images. This was demonstrated by showing some work by Akiyoshi Kitaoka’s Rotating Snakes (left image below) and Mondrian’s Broadway Boogie Woogie. It reminded me of the SIGGRAPH 2008 paper on
Self-Animating Images: Illusory Motion Using Repeated Asymmetric Patterns by Ming-Te Chi et al.
The images in this paper seem to use a similar principle and are very effective and conveying motion.
- Our central vision has high acuity whereas our peripheral vision was lower acuity. We dont seem to notice the fact that our peripheral vision has lower resolution since we move our eyes rapidly over scenes. This reminded me of the keynote from IEEE Vis 2004, where the speaker, Wilson Geisler, gave an amazing demonstration of this phenomena.
She said that Mona Lisa’s smile was particularly enigmatic since one feels her expression changing depending on whether one focuses on the eyes or the mouth. Her lab has conducted research on the same and by filtering if they have been able to create representations of how our peripheral and central vision interpret the painting, as shown here
The last part of the talk focused on the use of stereopsis by artists in making more realistic paintings. She found that many artists lack the ability to see stereopsis and that makes them see a flat world which they capture on canvas. This can be identified by looking at the glint in the eye of the photo/painting and if both the glints are not synchronized, chances are that the person lacks the ability to see stereopsis. Through her research, she found many famous artists to have misaligned eyes including the painter Rembrandt. Photographs of Babe Ruth too seemed to imply that he had misaligned eyes.
The end of the talk kept us wanting for more and I guess that implies that it was a great keynote talk.
He motivated the audience by making a very strong case for why there is a need to use Visual Analytics software. He basically said that there was a much wider customer base, than one would imagine, for quality visual analytics tool. What was interesting to me was that he said that his definition of a successful visual analytics tool was how widely adopted that tool was. I personally believe that its the best way to make sure that users have the power of visual analytics at their finger tips.
He basically said that we havent optimized the impact of visual analytics until you help users with their own data. The demo of Tableau was my favorite part, where he would end up interacting with simple datasets to show how easy it was to get insight or just know more about the data.
I think my favorite quote from the talk was ‘Visual analytics can help people test their hunches even when they lead to nowhere.’ :) This was great since this is exactly the purpose of visualization. The idea of interacting with your data to learn more but also just confirm what you already know.
He then showed an amazing demo of presidential donations data from new york city. Comparisons of Obama and McCain showed some wondreful, interesting patterns in parts of new york, like the upper east side and so on. Some of those patterns were expected for those who know the demographics of new york city.
Some of the highlights of the talk were:
- Most analytics tasks dont result in ‘Aha’ discoveries.
- People dont like to admit they need outside help to make discoveries about their own data.
- Visualization and Visual Analytics helps people think of what questions to ask. More importantly, it helps them enter the Visual analysis cycle of interact, explore, visualize, obtain insight – rinse and repeat :)
His most imprtant statement of the keynote speech was that “The number one reason people buy visual analytics software was to save time.”
I think it was the kind of keynote that makes you think and shakes you up a bit. I agree that we needed to hear some of those words. I particularly enjoyed the talk, since I had mentioned Tableau Software in one of my previous blog posts on ‘Visualizing companies leading the way.’ After listening to the talk, I feel more confident that Tableau will make a big difference in the Visual analytics and business analytics community. If you have anything to say about the keynote, please feel free to add a comment.
The IEEE Visweek 2008 had a bunch of wonderful workshops. I particularly enjoyed the ‘Understanding Federal Funding: Agencies, Initiatives, and Peer Review‘ workshop which was led by Terry Yoo who represented National Institutes of Health (NIH). The other workshop participants were Christine Chalk from the Department of Energy (DOE) and Larry Rosenblum from the National Science Foundation (NSF).
They basically said that given the state of the economy, it is almost certain that the funding situation is going to be bad for the next fiscal year. But given that, I really enjoyed the advice they had for the audience. Terry spoke about how its important to find the right institute within the NIH to apply to. He stressed on the fact that its extremely important to send your grant to the right place to ensure that you get funding for your research. Here’s a list of the institutes that comprise NIH. Christine spoke about DOE’s diverse funding opportunities but how sometimes they get less applications than they expect. Its definitely something to remember for all those who can apply for funding to the DOE.
Larry Rosenblum probably gave some priceless advice regarding how a grant should be written and what are the ‘donts’ about a grant. He gave some excellent examples of things that worked in grants. All their slides can be downloaded at http://erie.nlm.nih.gov/~yoo/edu/tutorials/vis2008/index.html
The next workshop that I was very excited about was the ‘From Theory to Practice: Design, Vision and Visualization‘ which was organized by Lyn Bartram, Maureen Stone and Diane Gromala. They have some slides online at http://www.stonesc.com/Vis08_Workshop/. I look forward to some more such endeavors from the community. We need many more such discussions to make sure that no one ever uses a rainbow colormap again :)
Some of the other interesting workshops that were conducted can be found at http://vis.computer.org/VisWeek2008/session/workshops.html. Unfortunately, I couldnt attend them but will definitely hope to look at some of the slides from those workshops. If any of you attended those workshops, pls feel free to comment on them in the comments section.
Geometry-Based Edge Clustering for Graph Visualization
WeiWei Cui, Hong Zhou, Huamin Qu, Pak Chung Wong, Xiaoming Li
This paper reminded me of a similar paper from Stanford called ‘Flow Map Layout‘. That paper had some crisp, clean images which conveyed information effectively. In the talk, WeiWei presented geometric handles that they provide to control the edge clustering for improved graph visualization. They use a ‘control mesh’ that can be generated at different levels of detail (manually or automatically) based on the graph structure. I thought the results looked good, but its use is limited if the user cannot easily interact with the provided tools.
On the Visualization of Social and other Scale-Free Networks
Yuntao Jia, Jared Hoberock, Michael Garland, John C. Hart
In this paper, they introduce a graph layout algorithm that reduces the overlap between edges.
This was interesting since later in the week, Frank van Ham in his talk said that users tried hard to reduce the number of overlapping edges when they were asked to make their own layout of social networks.
Exploration of Networks Using Overview+Detail with Constraint-based Cooperative Layout
Tim Dwyer, Kim Marriott, Falk Schreiber, Peter J. Stuckey, Michael Woodward, Michael Wybrow
This paper discussed the results of collaborations with biologists to visualize complex biological
networks. I think we need to encourage such collaborative projects and maybe have some more
initiatives in the form of awards and incentives to get more researchers to work in such a collaborative manner.
Rapid Graph Layout Using Space Filling Curves
Chris Muelder, Kwan-Liu Ma
In this talk, Chris presented an interesting idea to create graph layouts. In particular, I liked the use of the Peano curve (one of the space filling curves used in the paper) for graph layout. I think that semantic zooming will make their work extremely interesting and as per the future work section in their paper, they are planning to do it.
Improving the Readability of Clustered Social Networks using Node Duplication
Nathalie Henry, Anastasia Bezerianos, Jean-Daniel Fekete
This was an extremely enjoyable talk, particularly because the Nathalie did a very good job of conveying the content of the paper by mixing in some well time humor. Her talk focused on
the visualizing social networks and discussed some of her previous work and proposed results of an evaluation of some new techniques to visualizing such graphs. Duplicating an actor in social networks was found to be promising for social network analysis. Since analyzing social networks is such a topic of interest for security purposes, this work is quite timely.
Effectiveness of Animation in Trend Visualization
George Robertson, Roland Fernandez, Danyel Fisher, Bongshin Lee, John Stasko
This was an interesting presentation by George, where he presented some of their work on evaluating
the effectiveness of animations in presentations and analysis. Their user study particularly
was focused on evaluating the effectiveness of techniques such as those used in Hans Rosling’s
famous talk from TED 2006 on Global health, poverty and infant mortality etc. If you have not
seen the video yet, it can be seen at http://video.google.com/videoplay?docid=2670820702819322251
Perceptual Organisation in User-Generated Graph Layouts
Frank van Ham, Bernice E. Rogowitz
This was a very enjoyable talk since in this project, they took the approach of letting random web users create their own layout for social graphs. The idea was to analyze their layouts and see if they can uncover underlying clusters in the social graphs. They found that users used the edges to delineate the boundaries of the graph. This is completely contrary to all the graph layout algorithms which have a sunburst style layout. It was great to see Bernice Rogowitz at the conference again. She has done some wonderful work as regards color and its used in creating effective visualizations. Here’s a list of some of her publications. The ‘Which Blair‘ project from Vis ’01 and the ‘The end of the rainbow‘ work with Lloyd Treinish.
Spatially Ordered Treemaps
Jo Wood, Jason Dykes
This was a talk in which they used treemaps whose ordering was based on spatial constraints. It
was a project where the authors worked with geographers and found that they preferred to maintain
spatial relationships between items. The layout along with some bezier vectors that they added
to the layouts to show some temporal relationships made for a unique talk.
The InfoVis 2008 was an extremely enjoyable and interesting talk where Jake actually talked about some of the truth that visualization researchers need to here. It was very interesting to hear how he
focused on information-centricity and how focusing on the data and information is critical.
What most strongly resonated with me was the emphasis on “Visualization-based interaction”. He
said that visualizations needs to go beyond simply showing information to the user and start focusing on becoming a rich medium for exploration, annotation and sharing.
Most of the remaining talk focused on how his company had dealt with some of the challenges
of deploying a visualization software, Command post of the future (CPOF) where they needed to get real time data to military personnel for facilitate fast decision making and reduce the uncertainty for all the
parties involved. The ability to share and collaborate remotely made it a success and I think it is truly a fascinating visualization story.
He introduced a concept of collaboration literacy which extends the concept of visual literacy
and verbal literacy. Visual literacy and verbal literacy are well known. Collaboration literacy discusses how we collaborate and how we learn to collaborate, right from simple daily face-to-face meetings to virtual telepresence.
Some of the things that he mentioned that according to me are critical for us visualization researchers are:
- We need to realize that visualizations are the medium that users use to understand information.
- Dont make users do extra work. I completely agree with this and I think that Google’s success
has been partly due to this exact reason. For example, now when we try to search for keywords
in Google, they suggest commonly used keywords that will make it easier to search.
- More ways to incorporate self-synchronization where users across the board are able to understand
the information and are able to synchronize their knowledge with what other users have at that time.
Alternatively, some annotations and interactions can let one know that the user on the other
side has understand some key concepts and so they are on the same page and start interacting from there on.
As the November elections approach, I’ve been observing more and more websites, newspapers and even a few researchers conveying data and trends using visualization techniques.
EveryMomentnow.com is a wonderful website that tracks the number of news articles that are focused on one of the presidential candidates. Here is an example visualization from their main page. It reminded me of Tufte’s Sparklines, which convey information so succinctly.
But, New York Times scores a home-run again by making an interactive election visualization that allows readers to explore their data, create your own prediction map and much much more. It even allows you pick different ‘opinion polls’ and see what they’re predicting. Here’s a snapshot from the ‘Wall Street Journal/NBC News’ predictions. It can be found at http://elections.nytimes.com/2008/president/whos-ahead/key-states/map.html
Microsoft’s Live Labs too has developed “Political Streams” which is basically an integrated visualization tool of sorts, that allows one to visualize the attention an event is getting. They divide the ‘attention’ into ‘news’ attention and ‘blog’ attention. It would be more interesting if they would allow users to start their own streams.
Well, I’m happy to see that there has been some significant interest in the academic community as well.
Geoff Draper from the University of Utah will be presenting a paper in the IEEE Information Visualization 2008 conference (which is part of IEEE VisWeek 2008) titled: “Who Votes For What? A Visual Query Language for Opinion Data” by Geoffrey M. Draper & Richard F. Riesenfeld.
Chris Healey’s 2004 US Presidential Election research: http://www.csc.ncsu.edu/faculty/healey/US_election/
Jean Daniel Fekete’s 2004 French Election related work: http://www.lri.fr/~fekete/elections2004/index.en.html
As always, my favorite visualization website, Many Eyes, is full of users trying to visualize everything from
Many more such elections related visualization have been created and can be browsed at http://services.alphaworks.ibm.com/manyeyes/browse/visualizations?q=presidential
I feel that even though some visualizations provide interactivity, most of them seem to stay away from using some of the wonderful information visualization research that has been going on to allow users to obtain insight such as the ‘Visual Information Exploration for the Web‘ by Sheelagh Carpendale. I’m sure you have seen some good and bad visualizations this election season. Please feel free to post links or mention them in the comments.