Facets Dive

The ability to verify the ground truth of data distributions is critical in supervised learning. Supervised ML involves training datasets with labels. These labels constitute the target values. An ML algorithm will be trained to predict them. However, some or all of the labels might be wrong. The accuracy of the predictions might not be sufficient.

With Facets Dive, we can explore a large number of data points interactively and analyze their relationships.

Building the Facets Dive display code

We first import the display and HTML modules from IPython:

# Display the Dive visualization for the training data
from IPython.core.display import display, HTML

The next step is to convert a pandas DataFrame containing the training or testing data into JSON. You can run an example that was inserted in the notebook before continuing:

# @title Python to_json example {display-mode: "form"}
from IPython.core.display import display, HTML
jsonstr = train_data...