I need a description/explaination of how an (already implemented) algorithm works. It should be that easy that somebody who doesnt know much about neural networks also gets it. So you would have to start real low when explaining this. (but it doesnt have to get very deep, and not very long (2-3 pages), maybe drafts will be useful.
Further i need to adapt the algorithm:
i will only use the algorithm for incomplete data (Dataloader_incomplete).
i need it to return not only the loss, but also the complete Dataset with the computet values (as .csv file).
The algorithm is free available on GitHub:
[login to view URL]
There is also a paper where the idea of MRNN is introduced, but not very comprehensible and extensive enough for non-professionals.
So i need an explaination for:
- how these 2 neural networks work and how they are connected (there is one for interpolation, one for imputation), where one can work both forward and backward (how and why?)
- what for the multiple imputation in the dropout is and how dropout works