Day 10 - Using TensorBoard for Visualization

TensorBoard is an interactive visualization tool that can be used to view learning curves during training, compare them between multiple runs, visualize the computation graph of the network, etc.

Data to be visualized has to be output into special binary log files - event files.

Each binary data record is called a summary.

The TensorBoard server monitors the log directory and automatically updates the visualizations with live data (with a short delay).

To visualize data from multiple runs of program, point the server to a root log directory and configure program to write do a different subdirectory every time it runs.

Keras provides a TensorBoard() callback that creates the log directory and event files, and writes summaries to them.

TensorFlow also offers a lower-level API in the tf.summary package to log data that can be visualized using TensorBoard.

Use TensorBoard to visualize training and validation performance of network on Fashion MNIST