Within BBVA, we govern an active data science online community with regular exchanges of references being from academia, start-ups, GAFA, banking, discussing the latests research publications, model applications, toolkit release, data engineering deployments, trends in visual analytics and more. This is part of our culture to cross-pollinate practices and approaches within our different teams and countries. This year we saw many inspiring talks that we think should be spread widely. Here are a few that we hope you enjoy as much as we did:
Introduction to Gaussian Processes by Neil Lawrence
This talk gives a brilliant overview of regression and classification model (e.g. GAM, SVM, including some neural networks) that determine Gaussian Processes. It is mathematically robust with the capacity among other things to provide bayesian confidence intervals. Additionally, it gives a bayesian focus on machine learning with a capacity to process large amounts of data (we encourage you to read Gaussian Processes for Machine Learning).
Visual Analysis of Hidden State Dynamics in Recurrent Neural Networks by Hendrik Strobelt
Data visualizations are increasingly being used to break algorithmic black boxes. This line of work comes from the needs of data scientists and society in general to understand what models do and to be able to explain their implications. Currently we only know that they work. This presentation of LSTMVis shows a visual analysis tool for recurrent neural networks with a focus on understanding their hidden state dynamics.
Design as a Search Problem by Mike Bostock
In this fascinating talk Mike Bostock establishes design as a “search problem” and discusses some strategies for addressing this problem that can help make better and more efficient design data visualizations.
The Cost of Missing Something by Tricia Wang
Why do so many companies make such bad decisions? Tricia Wang demystifies the pitfalls of big data, with stories from Nokia to Netflix to the Oracles of Ancient Greece, and shares how integrating “thick data” can help companies avoid bad decisions and thrive in the unknown. Among other things, Tricia runs the resourceful Ethnography Matters.
GoodCityLife.org by Daniele Quercia
In his Databeers talk, Daniele Quercia presents a human approach of wayfinding using data analytics that suggests sensorial and emotional routes to go from A to B in a city. The aesthetic layer trained as part of a game and that exploits Google Street View imagery. Find out more about Smelly Maps, Happy Maps etc in Good City Life.
We hope you enjoyed these talks and hope to share more in 2017!