Differentiable Neural Computer
A paper from DeepMind proposing a differentiable neural computer architecture building on NTM, improving memory access mechanisms and supporting dynamic memory allocation and a link matrix. It is an important evolution of memory-augmented neural networks.
Timeline (last 90 days)
The concept of the neural memory module can be traced back to early work such as Neural Turing Machines and the Differentiable Neural Computer, whose core idea is to model memory read and write operations as differentiable computations.
Memory-Augmented Neural Networks (MANN) are a class of architectures that attach an external memory module beyond a standard neural network, traceable at the earliest to Neural Turing Machines and the Differentiable Neural Computer.
All Facts (2)
The concept of the neural memory module can be traced back to early work such as Neural Turing Machines and the Differentiable Neural Computer, whose core idea is to model memory read and write operations as differentiable computations.
50%UnverifiedMemory-Augmented Neural Networks (MANN) are a class of architectures that attach an external memory module beyond a standard neural network, traceable at the earliest to Neural Turing Machines and the Differentiable Neural Computer.
50%