My experiences at the VAST challenge
I have been fortunate to compete in the Visual Analytics Science and Technology Challenge or VAST Challenge as it is called, for two consecutive years now. Having been through the rigor of the competition, I felt I could share some experiences of working through it and throw some general light on what this competition is about.
About the challenge
The VAST challenge is an annual competition in Visual Analytics conducted by the Visual Analytics community, with the committee of judges hailing from University of Amherst, Massachusetts, the US Air Force Defense National Laboratory, etc. Datasets are released pertaining to specific scenarios which are termed as mini-challenges (MCs).
Data is released typically around April, with submissions due around mid-July. Submissions need to include a set of text answers, images that help narrate the story and a 4-minute video where participants can delve deeper into techniques and assumptions. A few examples here.
The technical stuff:
The theme of the challenge over the last 2 years was around how a mythical bird species had begun to decline in number at a wildlife preserve, and how various factors are believed to have been contributing to it. The premise was in the participants building a comprehensive narrative using the data to formulate a relevant hypotheses.
For e.g. in 2017, three datasets pertaining to traffic patterns, chemical sensor readings and satellite imagery were released. The insights from these three needed to be bundled in what was called the Grand Challenge, where participants uncover the big picture knitting the findings from the three MCs. In 2018, the challenge organizers decided to base the scenario from 2017, this time pitting participants with datasets that possibly point to contrary evidence to what they discovered in 2017. Three similar MCs were released in 2018, but with no Grand Challenge.
What I’ve gained?
It started as an in-class assignment in our Visual Analytics course at SMU. Our team of three chose to pick an MC each in 2017, so that we could submit the grand challenge. In 2018, the workload was more demanding as it needed more expertise around audio file processing and time series analysis. The experience of donning the hat of a visual detective where you arrive at patterns or anomalies in data, that help weave a story line is something that has spurred me on. Also, it is as important to communicate the story from the findings, rather than develop novel visualizations that lack a meaningful story underneath. This might be quite obvious, but have generally observed teams failing to balance out the two.
The IEEE VIS Conference
The IEEE VIS conference encompasses a wide variety of data visualization related events and is the most premier event in this arena. Hosted by IEEE and a wide variety of sponsors including Tableau, Microsoft Research, Disney, etc. this is where the work done towards the VAST challenge gets recognized and officially culminates. One gets to listen to what fellow award winners developed for their solutions. Also, it is a great opportunity to network with employers and pioneers in the data visualization community. Each year, the venue is different, which gave us the experience of visiting two countries, the USA and Germany. The VAST Challenge workshop takes place generally on the first day of the IEEE VIS of each year, which is a Sunday. The next one is coming up at Vancouver in 2019.
One of the most important piece is also the feedback the organizers take from the participants on what kind of data-sets they wish to work on in the upcoming year. Some of the themes discussed this year were on more deep learning oriented, object detection related scenarios. So, things are bound to get interesting in the coming years!
Why this might interest you?
For two consecutive years, I experienced someone asking me how a ‘management’ university has managed to compete in a visual analytics conference. I disagree that the word ‘management’ can weigh down what a university’s faculty and departments are capable of. For that matter, it is open to anyone from academia/industry. So, an interest in data visualization and the right approach towards the challenge theme can make anybody a worthy contender in this challenge. Also, the data is mostly well curated, and the challenge questions are well defined, and you do not have to spend that cliched 80% of your time in the ETL stuff.
Cheers,
Kishan
Links
Link to my talk in IEEE VIS 2017: https://vimeo.com/242504836
Link to my talk in IEEE VIS 2018: https://vimeo.com/319158072
VAST Challenge 2017: http://www.vacommunity.org/VAST+Challenge+2017
VAST Challenge 2018: http://www.vacommunity.org/VAST+Challenge+2018
The Visual Analytics Benchmark Repository, where one can find all submission files.
