Figure 2-Graphcoloring

[chengquan50]

The reward of the graphcoloring environment is an integer, and the convergent mean generated by the three random seeds needs to be divisible by 3, but 3 times the convergent mean given in the graph is obviously not an integer.

[sash-a]

Hi @chengquan50 👋

Not sure why the mean reward over 3 random seeds should be divisible by 3? The reward is simply -num_colours_used which may or may not be divisible by 3

[chengquan50]

My concern is that since reward is an integer in this environment, according to the legend, "Experiments were run with three different seeds, with the median represented as a blue curve and the min/max as the shaded region.", the blue curve should also be an integer, right? Thanks

[sash-a]

I see, it's possible that it was mean instead of median in this case by mistake, but I'm not sure. Maybe @clement-bonnet or @surana01 or @coyettev can give you a better answer as they worked on the graphs in the paper