What is a good explanation of how to read the histogram feature of TensorBoard?

Question

Question is simple, how do you read those graphs? I read their explanation and it doesn't make sense to me.

I was reading TensorFlow's newly updated readme file for TensorBoard and in it tries to explain what a "histogram" is. First, it clarifies that its not really a histogram:

Right now, its name is a bit of a misnomer, as it doesn't show histograms; instead, it shows some high-level statistics on a distribution.

I am trying to figure out what their description is actually trying to say.

Right now I am trying to parse the specific sentence:

Each line on the chart represents a percentile in the distribution over the data: for example, the bottom line shows how the minimum value has changed over time, and the line in the middle shows how the median has changed.

The first question I have is, what do they mean by "each line". There are horizontal axis and there are lines that make a square grid on the graph or maybe the plotted lines, themselves. Consider a screen shot from the TensorBoard example:

What are they referring to with "lines"? In the above example, what are the lines and percentiles that they are talking about?

Then the readme file tries to provide more detail with an example:

Reading from top to bottom, the lines have the following meaning: [maximum, 93%, 84%, 69%, 50%, 31%, 16%, 7%, minimum]

However, its unclear to me what they are talking about. What are lines and what percentiles?

It seems that they are trying to replace this in the future, but meanwhile, I am stuck with this. Can someone help me understand how to use this?

Answer 1

The lines that they are talking about are described below:

as for the meaning of percentile, basically, the 93rd percentile means that 93% of the values are situated below the 93rd percentile line.

Learn Tensorflow to have a better understanding. Since Tensorflow and Machine Learning Courses are quite interrelated, one can always go for the latter to have a better understanding.