Closed matt-gardner closed 8 years ago
All this documentation looks great - I think we could do with a top level description/ picture of how we have broken down the memory network architecture and how it makes it easy to implement lots of different architectures, e.g encoder > sentence encoder > knowledge selector > knowledge combiner > memory updater > entailment combiner etc. I'm happy to do this in the afternoon today.
This PR mostly just adds a bunch of READMEs and updates the example experiment configs. If you could check this for anything that is unclear, or could use more documentation, that'd be great.