#ai # [[Epistemic status]] #shower-thought # Neural networks ## Backpropagration complexity ![[Pasted image 20221116101055.png]] A good way to think about backpropagation is that its algorithmic complexity is $O(mn)$,$O(mn)$, where $m$ is the number of neurons in this layer and $n$ is the number of neurons in the next layer. One could also show that this is equivalent to $O(W)$,$O(W)$, where $W$ is the number of synapses in the network. # External links