Currently not implemented ([link](https://github.com/sinzlab/neuralpredictors/blob/main/neuralpredictors/layers/readouts/factorized.py#L180)), however, logic from Sinz 2019 could be borrowed to do it ([here](https://github.com/atlab/attorch/blob/886ba942938fc875718c0831544af12a0eaf02a9/attorch/layers.py#L690) and [here](https://github.com/sinzlab/Sinz2018_NIPS/blob/master/nips2018/architectures/base.py#L69))
