Neural Turing Machines

2014/10/20 by Alex Graves, Greg Wayne, Graves, Alex +3 · 11 voices · 89 citations
Computer Science · Engineering · #Advanced Memory and Neural Computing #Ferroelectric and Negative Capacitance Devices #Neural Networks and Applications #cs.NE

paper · pdf · doi:10.48550/arxiv.1410.5401

openalex publication_date 2014/10/20 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We extend the capabilities of neural networks by coupling them to external memory resources, which they can interact with by attentional processes. The combined system is analogous to a Turing Machine or Von Neumann architecture but is differentiable end-to-end, allowing it to be efficiently trained with gradient descent. Preliminary results demonstrate that Neural Turing Machines can infer simple algorithms such as copying, sorting, and associative recall from input and output examples.

Citations

• On the Computational Power of Neural Nets
• Sequence to Sequence Learning with Neural Networks
• Neural Machine Translation by Jointly Learning to Align and Translate

Cited by

• Short-Term-to-Long-Term Memory Transfer for Knowledge Graphs under Partial Observability
• Pretraining Recurrent Networks without Recurrence
• Coordinating from Memory: Graph-Structured Experience Reuse for Multi-Agent Adaptation in Dynamic Manufacturing
• Neuro-Symbolic Meta-Policies for Temporal Knowledge-Graph Memory under Partial Observability
• Exposure is not manifestation: measurement target and output resolution jointly determine which behavioural-faithfulness evaluator wins
• Candidate Attended Dialogue State Tracking Using BERT
• Program-as-Weights: A Programming Paradigm for Fuzzy Functions
• Phionyx: A Deterministic AI Runtime Architecture with Structured State Management and Pre-Response Governance
• Kinaema: a recurrent sequence model for memory and pose in motion
• Whither symbols in the era of advanced neural networks?
• Hierarchical Reasoning Model
• From Memories to Maps: Mechanisms of In-Context Reinforcement Learning in Transformers
• TALES: Text Adventure Learning Environment Suite
• Partially Observable Reinforcement Learning with Memory Traces
• General Intelligence Requires Reward-based Pretraining
• Efficient Parallel Samplers for Recurrent-Depth Models and Their Connection to Diffusion Language Models
• Lessons from Neuroscience for AI: How integrating Actions, Compositional Structure and Episodic Memory could enable Safe, Interpretable and Human-Like AI
• Compressing Observation History into Agent Memory: Distilling Transformers into Recurrent Transformers
• Memory Layer: Train the In-Model Cache for Recommendation Models
• Beyond Heuristics: A Decision-Theoretic Framework for Agent Memory Management
• Beyond Sliding Windows: Learning to Manage Memory in Non-Markovian Environments
• Embedded Safety-Aligned Intelligence via Differentiable Internal Alignment Embeddings
• Forgetful but Faithful: A Cognitive Memory Architecture and Benchmark for Privacy-Aware Generative Agents
• ARC-AGI Without Pretraining
• Memory-Integrated Reconfigurable Adapters: A Unified Framework for Settings with Multiple Tasks
• Exact Learning of Arithmetic with Differentiable Agents
• Goal-Directed Search Outperforms Goal-Agnostic Memory Compression in Long-Context Memory Tasks
• Weaver: Kronecker Product Approximations of Spatiotemporal Attention for Traffic Network Forecasting
• CellARC: Measuring Intelligence with Cellular Automata
• BudgetMem: Learning Selective Memory Policies for Cost-Efficient Long-Context Processing in Language Models
• Hydra: Dual Exponentiated Memory for Multivariate Time Series Analysis
• Metis: Memory Foundation Model
• Reactive Transformer (RxT) -- Stateful Real-Time Processing for Event-Driven Reactive Language Models
• EverMemOS: A Self-Organizing Memory Operating System for Structured Long-Horizon Reasoning
• Gradual Forgetting: Logarithmic Compression for Extending Transformer Context Windows
• Energy-Efficient Domain-Specific Artificial Intelligence Models and Agents: Pathways and Paradigms
• Generalizable Reasoning through Compositional Energy Minimization
• Capability Ceilings in Autoregressive Language Models: Empirical Evidence from Knowledge-Intensive Tasks
• From Masks to Worlds: A Hitchhiker's Guide to World Models
• To Infinity and Beyond: Tool-Use Unlocks Length Generalization in State Space Models
• Multi-Agent Design Assistant for the Simulation of Inertial Fusion Energy
• ProtoTopic: Prototypical Network for Few-Shot Medical Topic Modeling
• Refine Thought: A Test-Time Inference Method for Embedding Model Reasoning
• MeSH: Memory-as-State-Highways for Recursive Transformers
• Artificial Hippocampus Networks for Efficient Long-Context Modeling
• ELMUR: External Layer Memory with Update/Rewrite for Long-Horizon RL Problems
• Memory-Augmented Generative AI for Real-time Wireless Prediction in Dynamic Industrial Environments
• ToolMem: Enhancing Multimodal Agents with Learnable Tool Capability Memory
• RheOFormer: A generative transformer model for simulation of complex fluids and flows
• Bridging Reasoning to Learning: Unmasking Illusions using Complexity Out of Distribution Generalization
• On The Statistical Limits of Self-Improving Agents
• A Computational Perspective on NeuroAI and Synthetic Biological Intelligence
• When Can AI Models Explain Learning? Validity Criteria for AI as Cognitive Models in Education
• Statistical Advantage of Softmax Attention: Insights from Single-Location Regression
• Mamba Modulation: On the Length Generalization of Mamba
• Spontaneous symmetry breaking and Goldstone modes for deep information propagation
• On the Importance of Multistability for Horizon Generalization in Reinforcement Learning
• Limitations of Normalization in Attention Mechanism
• Cross-Attention is Half Explanation in Speech-to-Text Models
• From Mimicry to True Intelligence (TI) -- A New Paradigm for Artificial General Intelligence
• Symbolic Feedforward Networks for Probabilistic Finite Automata: Exact Simulation and Learnability
• LLM Architecture, Scaling Laws, and Economics: A Quick Summary
• Beyond Memorization: Extending Reasoning Depth with Recurrence, Memory and Test-Time Compute Scaling
• Neural Field Turing Machine: A Differentiable Spatial Computer
• Re:Frame -- Retrieving Experience From Associative Memory
• Revisiting associative recall in modern recurrent models
• Scaling Memory-Augmented Neural Networks with Sparse Reads and Writes
• Memory-Augmented Transformers: A Systematic Review from Neuroscience Principles to Enhanced Model Architectures
• Hebbian Memory-Augmented Recurrent Networks: Engram Neurons in Deep Learning
• MLC-Agent: Cognitive Model based on Memory-Learning Collaboration in LLM Empowered Agent Simulation Environment
• Learning neuro-symbolic convergent term rewriting systems
• Learning Program Behavioral Models from Synthesized Input–Output Pairs
• RUMAA: Repeat-Aware Unified Music Audio Analysis for Score-Performance Alignment, Transcription, and Mistake Detection
• Attention-Based Models for Speech Recognition
• Associative Long Short-Term Memory
• Doodle Your Keypoints: Sketch-Based Few-Shot Keypoint Detection
• MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent
• Zero-Shot Task Generalization with Multi-Task Deep Reinforcement\n Learning
• Long-Sequence Memory with Temporal Kernels and Dense Hopfield Functionals
• Can "consciousness" be observed from large language model (LLM) internal states? Dissecting LLM representations obtained from Theory of Mind test with Integrated Information Theory and Span Representation analysis
• THIRDEYE: Cue-Aware Monocular Depth Estimation via Brain-Inspired Multi-Stage Fusion
• Neural Turing machine [wikipedia]
• Recurrent neural network [wikipedia]
• Types of artificial neural networks [wikipedia]
• Connectionism [wikipedia]
• Cognitive architecture [wikipedia]
• Google DeepMind [wikipedia]
• Glossary of artificial intelligence [wikipedia]
• History of artificial neural networks [wikipedia]

Discussions

• Neural Turing Machines [pdf] [hn, 75 points, 11 comments]
• Neural Turing Machines [hn, 60 points, 7 comments]
• Neural Turing Machines (2014) [hn, 55 points, 27 comments]
• Neural Turing Machines by DeepMind [pdf] [hn, 8 points, 1 comments]
• I just read Neural Turing Machines arxiv.org/abs/1410.5401. It’s insightful, especially the first 9 pages including the memory “addressing mechanisms”. I wish I had read it years earlier as it filled [bsky, 3 points, 1 comments]
• The real deal is this old paper: arxiv.org/abs/1410.5401 [bsky, 3 points, 1 comments]
• Neural Turing Machines [lobsters, 3 points, 0 comments]
• Neural Turing Machines (DeepMind) [hn, 2 points, 0 comments]
• Neural Turing Machines (2014) [hn, 1 points, 0 comments]
• To be clear, I believe the interesting areas of modularity that warrant investigation are world modeling, reasoning, planning, and memory, generally. Bring back fuzzy memory modules! [6] [6] arxiv.org [bsky, 0 points, 0 comments]
• The key innovation of NTMs lies in their ability to learn algorithms and perform computations on a flexible, dynamic memory structure, which is not fixed beforehand. paper: arxiv.org/abs/1410.5401 [bsky, 0 points, 0 comments]

Related

• Memory and attention in deep learning
• Implementing Neural Turing Machines
• Neural Turing Machines: Convergence of Copy Tasks
• Reinforcement Learning Neural Turing Machines - Revised
• Structured Memory for Neural Turing Machines