Optimization - Amazon Web Services (AWS) Tutorial

- [Instructor] One of the foundational problems in deep learning is this concept of optimizing. And let's take a look at a few different scenarios here that explain this principle. When you're trying to solve for a problem, there are local minimum and global minimum. And if you get stuck in a local minimum, you're not going to get the ultimate best solution. So in this scenario here, if you're slowly descending down and you get stuck inside of this local minimum, then you won't eventually get to this global minimum down here and get the best possible solution, which reduces the error rate of the classification problem you're trying to solve, for example. One of the intuitions behind this is to think about walking down a mountain in the dark. You could get stuck in a lake somewhere halfway down the mountain. That's the local minimum. But if you get to the very bottom, then that would be the global minimum. Another way to…