Modern computational systems for data analysis now allow developers to create a wide range of interactive visualizations to expose and explore their datasets. One of the most common and powerful of these at present for Python and R is Bokeh, which contains thousands of interactive data visualizations that can be harnessed and deployed relatively easily (see https://docs.bokeh.org/en/latest/docs/gallery.html for a few examples of these). In the first part of this workshop, we will explore the ranges of interactive visualizations available in Bokeh, discuss best practices for choosing the most effective and intuitive methods for visualizing specific types of data, and walk through the steps needed to deploy these visualizations on the web so that others can use them. The second part of the workshop then turns to TensorSpace and methods for visualizing the data produced by neural networks (like Tensorflow), focusing on the ways that the data is rendered and transformed within each layer as it works toward generating the end result. With TensorSpace we can now visually interrogate the data I/O that the neural network uses and generates at each step of the process, providing a full picture of how the NN produced its final models and results.

Objectives:

At the completion of the Workshop, attendees will be able to:
Explore the varieties of interactive data visualizations provided within the Bokeh and rBokeh frameworks
Harness specific interactive visualizations for specific data analyses
Deploy interactive Bokeh visualizations via the web so others can use them
Understand the ways that models are built and results are generated by neural networks
Harness visual methods to expose and interrogate the choices made within the neural network at every step/level of the process

Level: Intermediate/Advanced

Duration: 2 hours

Prerequisites: All participants are expected to bring a laptop with a Mac, Linux, or Windows operating system on which they have administrative privileges. Having prior exposure to basic data analysis and R or Python programming is highly recommended (using the Jupyter Notebook in Anaconda3 is the current recommended framework for developing in Python).

Github repository: https://github.com/rcc-uchicago/vis-for-data-analysis-bokeh-tensorspace/

Advanced Interactive Visualizations for Data Analysis (Bokeh + TensorSpace)