This is the StatQuest video “Neural Networks Pt. 2: Backpropagation Main Ideas,” created by Josh Starmer and posted to his StatQuest channel in 2020. It follows his introduction to neural networks and tackles the algorithm that lets them actually learn from data.
Starmer walks through backpropagation slowly and concretely, showing how a network measures its error and then works backward to figure out how much each weight contributed, so those weights can be nudged in the direction that reduces the error. He grounds the chain rule of calculus in a small worked example, so that the abstract idea of “propagating the error backward” becomes a sequence of visible, understandable arithmetic steps rather than a formula to memorize.
For a beginner, this is one of the clearest explanations of the single most important training algorithm in deep learning. It complements the 3Blue1Brown backpropagation video and the foundational 1986 backpropagation paper, offering the same core idea at an unhurried, example-driven pace.