Recreate Jenson Shannon Divergence

I looked at Jenson Shannon divergence and tried to recreate it. It needs probability distributions to find the distance between them. The standard anomalies function of tfdv requires just the statistics and the scheme to find the distance. I wanted to know what are the parameters tensorflow is taking to find the distance? In other words What all data is it taking from statistics/schema to find the distance.

Please view this topic.