MANN

An architecture that augments a standard neural network with an external memory module, tracing back to Neural Turing Machines and the Differentiable Neural Computer. It accesses a scalable memory matrix through differentiable read and write operations, overcoming the limitation that knowledge is fixed within weights.