IEEE VIS 2017: Machine Learning, Diversity, Parties

October 13, 2017

(This article was originally published at eagereyes, and syndicated at StatsBlogs.)

I've ignored the major new topic this year so far: machine learning. Another new thing this year, though way overdue, was that we finally started to talk about diversity. And then there were the parties.

Machine Learning

Machine learning made a big showing this year, though I managed to miss most of the relevant talks and events. In addition to the best paper at VAST, there were also two workshops and a tutorial on the topic. The Visualization in Data Science workshop had an interesting panel discussing the question of when humans need to be in the loop (Hadley Wickham deftly summed it up at the end as “machines are good at some things, humans are good at other things”).

A relevant paper that I couldn’t fit anywhere else was LSTMVis: A Tool for Visual Analysis of Hidden State Dynamics in Recurrent Neural Networks by Hendrik Strobelt, Sebastian Gehrmann, Hanspeter Pfister, and Alexander Rush. They developed a tool for exploring a class of neural networks that are good at learning sequences. It runs in the browser and they have a variety of example inputs to try it out with.

Despite my lack of coverage of the machine learning papers and events, this topic was quite noticeable this year and will undoubtedly become a common theme at the conference.

Diversity Panel

At last year’s Death of SciVis panel, the issue of diversity (or rather its lack) in the visualization field came up in the discussion. This year, the Diversity in Visualization panel with Robert S. Laramee (Organizer), Rita Borgo, Vetria Byrd, Aviva Frank, Kelly Gaither, Ronald Metoyer, and Erica Yang set out to highlight and address the issue.

The presentations covered a variety of topics, from statistics about gender diversity (or rather, its lack) to power, and a few ways to improve things. Kelly Gaither talked about how the U.S. will soon be majority minority (meaning no particular race will have a majority), but minorities still only make up 7% of STEM graduates. Jobs, however, are in STEM. That’s clearly a problem that need to be addressed. Rita Borgo also looked at numbers, comparing gender diversity in the VIS and CHI organizing committees.