Deep Learning with Keras

Keras example — memory network for question answering

In this example, we will build a memory network for question answering. Memory networks are a specialized architecture that consist of a memory unit in addition to other learnable units, usually RNNs. Each input updates the memory state and the final output is computed by using the memory along with the output from the learnable unit. This architecture was suggested in 2014 via the paper (for more information refer to: Memory Networks, by J. Weston, S. Chopra, and A. Bordes, arXiv:1410.3916, 2014). A year later, another paper (for more information refer to: Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks, by J. Weston, arXiv:1502.05698, 2015) put forward the idea ...