Symbolic Reasoning with Differentiable Neural Comput

Recent breakthroughs demonstrate that neural networks are remarkably adept at sensory 8 processing1 and sequence2, 3 and reinforcement learning4. However, cognitive scientists and 9 neuroscientists have argued that neural networks are limited in their ability to define vari10 ables and data structures5–9, store data over long time scales without interference10, 11, and 11 manipulate it to solve tasks. Conventional computers, on the other hand, can easily be pro12 grammed to store and process large data structures in memory, but cannot learn to recognise 13 complex patterns. This work aims to combine the advantages of neural and computational 14 processing by providing a neural network with read-write access to an external memory. We 15 refer to the resulting architecture as a Differentiable Neural Computer (DNC). Memory access 16 is sparse, minimising interference among memoranda and enabling long-term storage12, 13, 17 and the entire system can be trained with gradient descent, allowing the network to learn how 18 to operate and organise the memory in a goal-directed manner. We demonstrate DNC’s abil19 ity to manipulate large data structures by applying it to a set of synthetic question-answering 20 tasks involving graphs, such as finding shortest paths and inferring missing links. We then 21 show that DNC can learn, based solely on behavioral reinforcement14, 15, to carry out com22 plex symbolic instructions in a game environment16. Taken together, these results suggest 23