Tensorboard Histogram Summary usage guide

Tensorboard function is very fancy;
If you do not know how to interpret the information:
    then the egg;

Main content: How to interpret the information use of histogram dashboard

When you want to see the distribution of values for a tensor during the training process, the statistics for their distribution can be displayed in the Tensorboard page by using the following statement:

Tf.summary.histogram (' Summary_name ', tensor)

Tensorboard has two tabs to view the results of the histogram: histograms and distributions, the former displaying statistical results in a straight form, which provides more abstract statistical information. Histogram

The histogram has two modes of display: Overlay and Offset:

The meaning of overlay is overlay, and the corresponding pattern is shown in the figure above.
To enlarge a chart:

Each axis has the following meanings: Horizontal axis: Value Vertical: The number of different lines represents different time/step. If a later line coincides with an earlier line, the line is drawn as an overlay. Hovering the mouse over a point will add a black display of three numbers, meaning: When the step xxx1, there are xxx2 elements of the value (approximately equal to) xxx3.

Offset mode:

Axis coordinate meaning: horizontal axis: Is still the value vertical: The amount of time/step to stop the mouse in the picture before it appears. Distribution Map

Distribution BOARD, still do not understand how effective interpretation, temporarily skip it.