PR #24: Implementation for issue #32

[harveyvn]

• Setup 2 experiments to generate the data for Case 0 and Case 6.
• Visualize the repetitions with Violin Plot.
• Commit the data of Case 0 and 6 in both Random and OpO evolution.

Random Visualization:

• Case 0:
• Case 6:

One-plus-One Visualization:

• Case 0
• Case 6

[alessiogambi]

Uhm.. What I cannot understand is for Case 0. Is the horizontal line the optimum that one can reach? If so, why do we need to repeat the experiment 30 times as by definition nobody can do better than that?

[harveyvn]

Uhm.. What I cannot understand is for Case 0. Is the horizontal line the optimum that one can reach? If so, why do we need to repeat the experiment 30 times as by definition nobody can do better than that?

At the beginning, we don't know exactly whether the original scenario can reach the optimum value or not until we execute its experiment. So by repeating the experiment several times in both OpO and Random Evolution, I suppose some cases can be categorized a perfect case generated by AC3R. In such cases, we can add a condition to determine how easily the scenario reaches the maximum value and stop its running after 3-5 epochs in order to save time and resources. Correct me if I'm wrong.