Some Other Sides of Visualization in Support of Data Science
Often times when we think of data visualization in data science, we think of exploratory analysis on our raw data or explanatory analysis on our models or their output. However, visualization at UA is also addressing other elements in the data science pipeline, such as data engineering, debugging, and wrangling. In particular, I will present our work in the development of a distributed array toolkit, Phylanx, that transforms and executes NumPy programs in asynchronous distributed C++. Here we create visualizations to understand the emergent system behavior and to debug and optimize code. As part of this project, we have developed Roundtrip, a library for easing the movement of data within Jupyter Notebooks for interactive visualization. I will also discuss our ongoing study into how people from different domains think about the structure of their data, with a "cute & cuddly" call for participation at the end.