INFOCOMP-2017 Venice
--D. Thiebaut (talk) 07:17, 26 June 2017 (EDT)
INFOCOMP17Venice, Italy, June 26-19, 2017
I'm attending INFOCOMP17, a conference where I have already presented several papers in the past. I presented the paper detailing the work done by Kyra Gan, Class of 2017, on the data visualization of the spread of a virus. The paper is available here. Kyra is now (as of June 2017) a graduate student at CMU's Tepper School of Business.
One quote I heard at a panel on Tuesday expressed quite a bit my recent exploration of Neural Nets (NN), and in particular the use of TensorFlow and Long-Short-Term-Memory based network to predict time series: "Neural Nets are an art." It was said by Alexander Troussov, of IBM, Dublin. Getting NN to converge requires tweaking, and a lot of trials, playing with the number of layers, the width of layers, the type of cells, the output functions, etc.
I was also happy to meet colleagues from Wellesley College whose two daughters, Artemis and Danaë, are or have attended Smith, in different ways (regular student and during a junior-semester-abroad): Takis Metaxas, and Stella Kakavoulis. Takis presented a paper titled "Manipulation of Search Engine Results during the 2016 US Congressional Elections," which covers how the algorithms used by various search engines to return query results had been compromised, and tricked. Very interesting and a-propos research.
