{"url":"/method/neural-turing-machine","slug":"neural-turing-machine","name":"Neural Turing Machine","full_name":"Neural Turing Machine","full_name_withheld":false,"description_markdown":"A **Neural Turing Machine** is a working memory neural network model. It couples a neural network architecture with external memory resources. The whole architecture is differentiable end-to-end with gradient descent. The models can infer tasks such as copying, sorting and associative recall.\r\n\r\nA Neural Turing Machine (NTM) architecture contains two basic components: a neural\r\nnetwork controller and a memory bank. The Figure presents a high-level diagram of the NTM\r\narchitecture. Like most neural networks, the controller interacts with the external world via\r\ninput and output vectors. Unlike a standard network, it also interacts with a memory matrix\r\nusing selective read and write operations. By analogy to the Turing machine we refer to the\r\nnetwork outputs that parameterise these operations as “heads.”\r\n\r\nEvery component of the architecture is differentiable. This is achieved by defining 'blurry' read and write operations that interact to a greater or lesser degree with all the elements in memory (rather\r\nthan addressing a single element, as in a normal Turing machine or digital computer). The\r\ndegree of blurriness is determined by an attentional “focus” mechanism that constrains each\r\nread and write operation to interact with a small portion of the memory, while ignoring the\r\nrest. Because interaction with the memory is highly sparse, the NTM is biased towards\r\nstoring data without interference. The memory location brought into attentional focus is\r\ndetermined by specialised outputs emitted by the heads. These outputs define a normalised\r\nweighting over the rows in the memory matrix (referred to as memory “locations”). Each\r\nweighting, one per read or write head, defines the degree to which the head reads or writes\r\nat each location. A head can thereby attend sharply to the memory at a single location or\r\nweakly to the memory at many locations","description_state":"present","introduced_year":null,"introduced_by":{"title":"Neural Turing Machines","paper":"/paper/neural-turing-machines","first_author":"Alex Graves","n_authors":3,"url_abs":null,"archive_paper_url":"https://paperswithcode.com/paper/neural-turing-machines"},"source":{"url":"http://arxiv.org/abs/1410.5401v2","title":"Neural Turing Machines","url_on_a_paper_host":true},"code_snippet_url":"https://github.com/carpedm20/NTM-tensorflow","code_snippet_url_on_a_code_host":true,"categories":[{"area":"Sequential","area_id":"sequential","collection":"Recurrent Neural Networks","url":"/methods/category/recurrent-neural-networks","pwc_aliases":[]},{"area":"General","area_id":"general","collection":"Working Memory Models","url":"/methods/category/working-memory-models","pwc_aliases":[]}],"n_papers_tagged":22,"archive_num_papers":22,"papers_newest_first":[{"paper":null,"title":"Intelligent DoS and DDoS Detection: A Hybrid GRU-NTM Approach to Network Security","date":"2025-04-10","arxiv_id":"2504.07478","n_code_links":0,"syntology":null},{"paper":"/paper/memory-augmented-conformer-for-improved-end","title":"Memory-augmented conformer for improved end-to-end long-form ASR","date":"2023-09-22","arxiv_id":"2309.13029","n_code_links":1,"syntology":null},{"paper":null,"title":"FashionNTM: Multi-turn Fashion Image Retrieval via Cascaded Memory","date":"2023-08-20","arxiv_id":"2308.10170","n_code_links":0,"syntology":null},{"paper":"/paper/token-turing-machines","title":"Token Turing Machines","date":"2022-11-16","arxiv_id":"2211.09119","n_code_links":1,"syntology":null},{"paper":null,"title":"Unsupervised Speaker Adaptation using Attention-based Speaker Memory for End-to-End ASR","date":"2020-02-14","arxiv_id":"2002.06165","n_code_links":0,"syntology":null},{"paper":null,"title":"Memory-Augmented Recurrent Networks for Dialogue Coherence","date":"2019-10-16","arxiv_id":"1910.10487","n_code_links":0,"syntology":null},{"paper":null,"title":"A Neural Turing~Machine for Conditional Transition Graph Modeling","date":"2019-07-15","arxiv_id":"1907.06432","n_code_links":0,"syntology":null},{"paper":null,"title":"Understanding Memory Modules on Learning Simple Algorithms","date":"2019-07-01","arxiv_id":"1907.00820","n_code_links":0,"syntology":null},{"paper":null,"title":"A review on Neural Turing Machine","date":"2019-04-10","arxiv_id":"1904.05061","n_code_links":0,"syntology":null},{"paper":"/paper/few-shot-generalization-across-dialogue-tasks","title":"Few-Shot Generalization Across Dialogue Tasks","date":"2018-11-28","arxiv_id":"1811.11707","n_code_links":2,"syntology":null},{"paper":null,"title":"Context-Aware Neural Model for Temporal Information Extraction","date":"2018-07-01","arxiv_id":null,"n_code_links":0,"syntology":null},{"paper":null,"title":"A Taxonomy for Neural Memory Networks","date":"2018-05-01","arxiv_id":"1805.00327","n_code_links":0,"syntology":null},{"paper":null,"title":"Meta-Learning via Feature-Label Memory Network","date":"2017-10-19","arxiv_id":"1710.07110","n_code_links":0,"syntology":null},{"paper":null,"title":"Attention-Set based Metric Learning for Video Face Recognition","date":"2017-04-12","arxiv_id":"1704.03805","n_code_links":0,"syntology":null},{"paper":"/paper/tracking-the-world-state-with-recurrent","title":"Tracking the World State with Recurrent Entity Networks","date":"2016-12-12","arxiv_id":"1612.03969","n_code_links":5,"syntology":{"ran":0,"of":3,"unverified":3,"pointer_only":0}},{"paper":null,"title":"Neural Turing Machines: Convergence of Copy Tasks","date":"2016-12-07","arxiv_id":"1612.02336","n_code_links":0,"syntology":null},{"paper":"/paper/dynamic-neural-turing-machine-with-soft-and","title":"Dynamic Neural Turing Machine with Soft and Hard Addressing Schemes","date":"2016-06-30","arxiv_id":"1607.00036","n_code_links":0,"syntology":null},{"paper":null,"title":"Lie Access Neural Turing Machine","date":"2016-02-28","arxiv_id":"1602.08671","n_code_links":0,"syntology":null},{"paper":null,"title":"Empirical Study on Deep Learning Models for Question Answering","date":"2015-10-26","arxiv_id":"1510.07526","n_code_links":0,"syntology":null},{"paper":null,"title":"A Deep Memory-based Architecture for Sequence-to-Sequence Learning","date":"2015-06-22","arxiv_id":"1506.06442","n_code_links":0,"syntology":null},{"paper":"/paper/reinforcement-learning-neural-turing-machines","title":"Reinforcement Learning Neural Turing Machines - Revised","date":"2015-05-04","arxiv_id":"1505.00521","n_code_links":1,"syntology":null},{"paper":"/paper/neural-turing-machines","title":"Neural Turing Machines","date":"2014-10-20","arxiv_id":"1410.5401","n_code_links":34,"syntology":{"ran":4,"of":8,"unverified":4,"pointer_only":2}}],"papers_shown":22,"tasks":[{"task":"/task/question-answering","name":"Question Answering","papers":4},{"task":"/task/automatic-speech-recognition-2","name":"Automatic Speech Recognition","papers":2},{"task":"/task/automatic-speech-recognition","name":"Automatic Speech Recognition (ASR)","papers":2},{"task":"/task/machine-learning","name":"BIG-bench Machine Learning","papers":2},{"task":"/task/decoder","name":"Decoder","papers":2},{"task":"/task/information-retrieval","name":"Information Retrieval","papers":2},{"task":"/task/machine-translation","name":"Machine Translation","papers":2},{"task":"/task/retrieval","name":"Retrieval","papers":2},{"task":"/task/speech-recognition","name":"Speech Recognition","papers":2},{"task":"/task/translation","name":"Translation","papers":2},{"task":"/task/speech-recognition-1","name":"speech-recognition","papers":2},{"task":"/task/action-detection","name":"Action Detection","papers":1},{"task":"/task/activity-detection","name":"Activity Detection","papers":1},{"task":"/task/deep-learning","name":"Deep Learning","papers":1},{"task":"/task/dimensionality-reduction","name":"Dimensionality Reduction","papers":1},{"task":"/task/face-recognition","name":"Face Recognition","papers":1},{"task":"/task/form","name":"Form","papers":1},{"task":"/task/image-retrieval","name":"Image Retrieval","papers":1},{"task":"/task/intrusion-detection","name":"Intrusion Detection","papers":1},{"task":"/task/knowledge-graphs","name":"Knowledge Graphs","papers":1}],"tasks_shown":20,"n_tasks":32,"usage_by_year":[{"year":"2014","papers":1},{"year":"2015","papers":3},{"year":"2016","papers":4},{"year":"2017","papers":2},{"year":"2018","papers":3},{"year":"2019","papers":4},{"year":"2020","papers":1},{"year":"2022","papers":1},{"year":"2023","papers":2},{"year":"2025","papers":1}],"row_source":"methods_table","archive":{"source":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","archive_url":"https://paperswithcode.com/method/neural-turing-machine"},"syntology_read_at":"2026-09-24T18:15:14+00:00"}