Uploaded February 2023 | Updated November 2026, 3 hours ago
Training neural networks happens in a very straightforward and predictable way: nothing but a slight variation on gradient descent. The difficulty is that for complex models like neural nets, working out the gradient is not trivial. Instead of doing it by hand, like we've done so far, we use an algorithm to automate the process: backpropagation.
slides: mlvu.github.io/lecture06
lecturer: Peter Bloem
Training neural networks happens in a very straightforward and predictable way: nothing but a slight variation on gradient descent. The difficulty is that for complex models like neural nets, working out the gradient is not trivial. Instead of doing it by hand, like we've done so far, we use an algorithm to automate the process: backpropagation.
slides: mlvu.github.io/lecture06
lecturer: Peter Bloem










