Separating the 'what' and 'how' of compositional computation to enable reuse and continual learning

Open in new window