Reinforcement Learning Neural Turing Machines - Revised
arXiv:1505.00521
Abstract
The Neural Turing Machine (NTM) is more expressive than all previously considered models because of its external memory. It can be viewed as a broader effort to use abstract external Interfaces and to learn a parametric model that interacts with them. The capabilities of a model can be extended by providing it with proper Interfaces that interact with the world. These external Interfaces include memory, a database, a search engine, or a piece of software such as a theorem verifier. Some of these Interfaces are provided by the developers of the model. However, many important existing Interfaces, such as databases and search engines, are discrete. We examine feasibility of learning models to interact with discrete Interfaces. We investigate the following discrete Interfaces: a memory Tape, an input Tape, and an output Tape. We use a Reinforcement Learning algorithm to train a neural network that interacts with such Interfaces to solve simple algorithmic tasks. Our Interfaces are expressive enough to make our model Turing complete.
References in corpus (5)
Cited by in corpus (73)
- Improved Image Captioning via Policy Gradient optimization of SPIDEr
- Deep Reinforcement Learning for Dialogue Generation
- Go for a Walk and Arrive at the Answer: Reasoning Over Paths in Knowledge Bases using Reinforcement Learning
- Neural Programmer-Interpreters
- A Simple, Fast Diverse Decoding Algorithm for Neural Generation
- Natural Language Processing Advancements By Deep Learning: A Survey
- Learning to Optimize
- Variational inference for Monte Carlo objectives
- A deep learning approach to cosmological dark energy models
- Backpropagation through the Void: Optimizing control variates for black-box gradient estimation
- Neural Abstractive Text Summarization with Sequence-to-Sequence Models
- Autocurricula and the Emergence of Innovation from Social Interaction: A Manifesto for Multi-Agent Intelligence Research
- End-to-End Answer Chunk Extraction and Ranking for Reading Comprehension
- Neural Symbolic Machines: Learning Semantic Parsers on Freebase with Weak Supervision
- End-to-end Learning of Action Detection from Frame Glimpses in Videos
- Towards End-to-End Reinforcement Learning of Dialogue Agents for Information Access
- A Neural Transducer
- Monotonic Chunkwise Attention
- Memory-Augmented Recurrent Neural Networks Can Learn Generalized Dyck Languages
- MuProp: Unbiased Backpropagation for Stochastic Neural Networks
- Adaptive Neural Compilation
- An Attentional Neural Conversation Model with Improved Specificity
- Learning Efficient Algorithms with Hierarchical Attentive Memory
- Improving the Neural GPU Architecture for Algorithm Learning
- Controllable Video Captioning with POS Sequence Guidance Based on Gated Fusion Network
- Learning Simple Algorithms from Examples
- Video Captioning via Hierarchical Reinforcement Learning
- Review of end-to-end speech synthesis technology based on deep learning
- Variational Memory Addressing in Generative Models
- Strong Generalization and Efficiency in Neural Programs
- Data Distillation for Controlling Specificity in Dialogue Generation
- An Introduction to Deep Learning for the Physical Layer
- Neural Shuffle-Exchange Networks -- Sequence Processing in O(n log n) Time
- Semantic Parsing with Syntax- and Table-Aware SQL Generation
- A Framework for Searching for General Artificial Intelligence
- CDL: Curriculum Dual Learning for Emotion-Controllable Response Generation
- Lie Access Neural Turing Machine
- Extensions and Limitations of the Neural GPU
- Learning Online Alignments with Continuous Rewards Policy Gradient
- Incremental Text to Speech for Neural Sequence-to-Sequence Models using Reinforcement Learning
- How deep learning works --The geometry of deep learning
- Disentangled Representations in Neural Models
- Compositional Generalization via Neural-Symbolic Stack Machines
- Advances in Natural Language Question Answering: A Review
- Robust Sequence-to-Sequence Acoustic Modeling with Stepwise Monotonic Attention for Neural TTS
- Neural Symbolic Machines: Learning Semantic Parsers on Freebase with Weak Supervision (Short Version)
- Integrating Episodic Memory into a Reinforcement Learning Agent using Reservoir Sampling
- #HashtagWars: Learning a Sense of Humor
- Source-Critical Reinforcement Learning for Transferring Spoken Language Understanding to a New Language
- Reconstruct and Represent Video Contents for Captioning via Reinforcement Learning
- Simplified Stochastic Feedforward Neural Networks
- Neural Status Registers
- An online sequence-to-sequence model for noisy speech recognition
- Motion Planning for Heterogeneous Unmanned Systems under Partial Observation from UAV
- Finding a Needle in a Haystack: Tiny Flying Object Detection in 4K Videos using a Joint Detection-and-Tracking Approach
- Towards one-shot learning for rare-word translation with external experts
- Adaptive Correlated Monte Carlo for Contextual Categorical Sequence Generation
- Video Captioning with Text-based Dynamic Attention and Step-by-Step Learning
- VERAM: View-Enhanced Recurrent Attention Model for 3D Shape Classification
- Recursive Sketches for Modular Deep Learning
- SafeRoute: Learning to Navigate Streets Safely in an Urban Environment
- Generalized Coarse-to-Fine Visual Recognition with Progressive Training
- An Empirical Comparison of Syllabuses for Curriculum Learning
- ReGen: Reinforcement Learning for Text and Knowledge Base Generation using Pretrained Language Models
- Deep Algorithmic Question Answering: Towards a Compositionally Hybrid AI for Algorithmic Reasoning
- Enhancing Reinforcement Learning with discrete interfaces to learn the Dyck Language
- Less is More: Data-Efficient Complex Question Answering over Knowledge Bases
- DORB: Dynamically Optimizing Multiple Rewards with Bandits
- Memory networks for consumer protection:unfairness exposed
- (Yet) Another Theoretical Model of Thinking
- Formal Fields: A Framework to Automate Code Generation Across Domains
- Partially Non-Recurrent Controllers for Memory-Augmented Neural Networks
- General AI Challenge - Round One: Gradual Learning