Structured Memory for Neural Turing Machines

Abstract

Neural Turing Machines (NTM) contain memory component that simulates "working memory" in the brain to store and retrieve information to ease simple algorithms learning. So far, only linearly organized memory is proposed, and during experiments, we observed that the model does not always converge, and overfits easily when handling certain tasks. We think memory component is key to some faulty behaviors of NTM, and better organization of memory component could help fight those problems. In this paper, we propose several different structures of memory for NTM, and we proved in experiments that two of our proposed structured-memory NTMs could lead to better convergence, in term of speed and prediction accuracy on copy task and associative recall task as in (Graves et al. 2014).

4 pages, accepted to Reasoning, Attention, Memory (RAM) NIPS 2015 Workshop

Cited by in corpus (6)

• End-to-End Answer Chunk Extraction and Ranking for Reading Comprehension
• Dynamic Neural Turing Machine with Soft and Hard Addressing Schemes
• Learning to update Auto-associative Memory in Recurrent Neural Networks for Improving Sequence Memorization
• Learning Numeracy: Binary Arithmetic with Neural Turing Machines
• Neural Turing Machines: Convergence of Copy Tasks
• Memory-Augmented Recurrent Networks for Dialogue Coherence