GitLab

Summary

Tensorboard is a data visualization tool by Google mostly used for machine learning but may have some interesting use cases for us. But is it worth adding as a new dependency?

Description

Currently, we only use Matplotlib for data visualization since our analysis is mostly limited to static (histograms). But in the future, we may need more visualization power where Tensorboard's ability to scale and smooth, overlay and dynamically refresh may come in handy. It currently supports scalars, graphs (tensors), audio and image data whereas Matplotlib has a more comprehensive plotting arsenal as long as one sticks to static analysis.

One interesting use case would be to use both in conjunction, utilizing Tensorboard's dashboard as a UI with Matplotlib plots.