Finding Policies - Georgia Tech - Machine Learning

Watch on Udacity: www.udacity.co...
Check out the full Advanced Operating Systems course for free at: www.udacity.co...
Georgia Tech online Master's program: www.udacity.co...

[@michelaka6836]

Not only will explained, but immensely important

[@jhk921]

So, a state that was falsely assigned with a negative utility value will slowly move toward its true utility value because it will be edited by its true reward R(s) and its utility value in regard to its neighbors, and the same would hold for the others as well?

[@jhk921]

So, I can start with flat out 0s for the utility values, and this thing would still work? I think it would, and I'm pretty sure, but I've never actually put this algorithm to the test so I'm not really certain.

[@sunilrathee2479]

How to make max differentiable?

[@oldcowbb]

LogSumExp

[@fgfanta]

You have already used "t" for time so far, it is misleading to use it for the number of iteration in the algorithm; you could use, say, "i" instead