1. 1

From the paper itself: “We demonstrated both theoretically and empirically, that automatic use of MLE without justification may result in highly sub-optimal performance. Therefore, we alert applied scientists to the fact that even if the MLE seems to have taken into consideration all the prior knowledge about the problem (such as the CL algorithm), it still might be significantly improved upon. This effect is far wider than we had the space to demonstrate here, and cognizance of it may lead to significant performance boosts in various disciplines.”