My neural network was built and is being built, using third party packages like GoLearn
twitter: @Arturo0911
See here for GoLearn.
Data are loaded in as Instances. You can then perform matrix like operations on them, and pass them to estimators. GoLearn implements the scikit-learn interface of Fit/Predict, so you can easily swap out estimators for trial and error. GoLearn also includes helper functions for data, like cross validation, and train and test splitting.
GoLearn comes with practical examples. Dive in and see what is going on.
cd $GOPATH/src/github.com/Artur0911/NN/examples/knnclassifier
go run knnclassifier_iris.go