@@ -88,7 +88,7 @@ backPropagate :: forall shapes layers.
8888```
8989which takes a network, appropriate input and target data, and returns the
9090back propagated gradients for the network. The shapes of the gradients are
91−appropriate for each layer, and may be trivial for layers like `Rulu` which
91+appropriate for each layer, and may be trivial for layers like `Relu` which
9292have no learnable parameters.
9393
9494The gradients however can always be applied, yielding a new (hopefully better)