I think a nice, short definition of Slack would be, “A small gradient of the cost function in the vicinity of the optimum.” In other words, the penalty for making sub-optimal decisions is not very high, because the region of acceptable behaviors around the best one is fairly large.
This doesn’t seem right to me. I think you could describe slack formally in terms of machine learning, but I don’t think this is a correct description of it. I was going to propose a better one, but I don’t know what I think it is yet.
This doesn’t seem right to me. I think you could describe slack formally in terms of machine learning, but I don’t think this is a correct description of it. I was going to propose a better one, but I don’t know what I think it is yet.