what is the difference between backward and backwardloss in " Define a Custom Regression Output Layer " ?

In the demo of " Define a Custom Regression Output Layer " ( https://ww2.mathworks.cn/help/nnet/ug/define-custom-regression-output-layer.html ) , the method "backwardloss " or "forwardloss" is used, but in the example "Define a Custom Deep Learning Layer with Learnable Parameters" (https://ww2.mathworks.cn/help/nnet/ug/define-custom-deep-learning-layer.html), the name changed to be "backward" or "forward". so, what is the difference between backward and backwardloss? and when to invoke these functions?