Designing Neural Network Architectures using Reinforcement Learning
Explore this paper's citation graph
Summary
MetaQNN is introduced, a meta-modeling algorithm based on reinforcement learning to automatically generate high-performing CNN architectures for a given learning task that beat existing networks designed with the same layer types and are competitive against the state-of-the-art methods that use more complex layer types.
- Type
- article
- Published
- 2016-11-04
- Cited by
- 1,556
- References
- 44
- Access
- Open access
- OpenAlex
- https://openalex.org/W2556833785
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:1740355
Keywords
Reinforcement learning, Computer science, Pooling, Artificial intelligence, Convolutional neural network
References
- Regularization of Neural Networks using DropConnect
- Convex Optimization Algorithms
- Understanding the difficulty of training deep feedforward neural networks
- Reinforcement Learning for Robots Using Neural Networks
- Combinations of genetic algorithms and neural networks: a survey of the state of the art
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Generative NeuroEvolution for Deep Learning
- Recurrent convolutional neural network for object recognition
- Experience Replay for Real-Time Reinforcement Learning Control
- Learning from delayed rewards
- Algorithms for Hyper-Parameter Optimization
- Reinforcement Learning: A Survey
- Evolving Neural Networks through Augmenting Topologies
- Multi-Task Bayesian Optimization
- Making a Science of Model Search: Hyperparameter Optimization in Hundreds of Dimensions for Vision Architectures
- Practical Bayesian Optimization of Machine Learning Algorithms
- Self-improving reactive agents based on reinforcement learning, planning and teaching
- A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation
- Human-level control through deep reinforcement learning
- Caffe: Convolutional Architecture for Fast Feature Embedding
Cited by
- Deep Reinforcement Learning: An Overview
- Entropy Non-increasing Games for the Improvement of Dataflow Programming
- Hemingway: Modeling Distributed Optimization Algorithms
- Large-Scale Evolution of Image Classifiers
- A genetic programming approach to designing convolutional neural network architectures
- DeepArchitect: Automatically Designing and Training Deep Architectures
- A Framework for Designing the Architectures of Deep Convolutional Neural Networks
- Building effective deep neural network architectures one feature at a time
- Practical Neural Network Performance Prediction for Early Stopping
- Neural Optimizer Search with Reinforcement Learning
- Data-Driven Sparse Structure Selection for Deep Neural Networks
- Reinforcement Learning for Architecture Search by Network Transformation
- Learning Transferable Architectures for Scalable Image Recognition
- Analysis and Optimization of Convolutional Neural Network Architectures
- Practical Network Blocks Design with Q-Learning
- Squeeze-and-Excitation Networks
- Evolution of Convolutional Highway Networks
- Stochastic Gradient Descent Performs Variational Inference, Converges to Limit Cycles for Deep Networks
- Transfer Learning to Learn with Multitask Neural Model Search
- Hierarchical Representations for Efficient Architecture Search