A NTM consists of a memory bank and a controller network. The controller network, consisting of LSTM and MLP, controls read/write heads by focusing their attention softly, using a distribution over all memory addresses
A PyTorch tutorial for implementation of Dynamic Memory Network Plus for Visual and Textual Question Answering. The architecture employs unique attention layers that maintain a fixed sequential order with the help of an episodic memory module.
An end-to-end model for jointly learning the scene and facial features of an image for group-level emotion recognition