What is Backpropagation?

Backpropagation applies the chain rule to propagate error gradients backwards through a network's layers, producing the per-parameter adjustments that learning consists of. It has underpinned neural network training since the 1980s and remains the mechanism behind every model in current use, automated now by differentiation libraries that hide the calculation entirely. For an organization consuming models, it is context rather than a decision point — no governance choice depends on it. What does depend on governance is everything after training: which model is called, what it is permitted to read, what it is instructed to do and what is retained afterwards. Centralpoint treats the trained model as a service selected at runtime and concentrates on that layer, which is also why a change of model or provider requires no re-indexing and no rewriting of the organization's rules.