This node denormalizes the input data according to the normalization parameters as given in the PMML model input (typically coming from the Normalizer node). The affine transformation is therefore inverted and the original values recreated.
This node is typically used after test data is normalized and possible other learning/predictions be transformed back to the original range.