Designing Neural Network Architectures using Reinforcement Learning

• Main Authors: Baker, Bowen, Gupta, Otkrist, Naik, Nikhil, Raskar, Ramesh
• Format: Journal Article
• Language: English
• Published: 07.11.2016
• Subjects: Computer Science - Learning
• DOI: 10.48550/arxiv.1611.02167

At present, designing convolutional neural network (CNN) architectures requires both human expertise and labor. New architectures are handcrafted by careful experimentation or modified from a handful of existing networks. We introduce MetaQNN, a meta-modeling algorithm based on reinforcement learning to automatically generate high-performing CNN architectures for a given learning task. The learning agent is trained to sequentially choose CNN layers using$Q$ -learning with an$\epsilon$ -greedy exploration strategy and experience replay. The agent explores a large but finite space of possible architectures and iteratively discovers designs with improved performance on the learning task. On image classification benchmarks, the agent-designed networks (consisting of only standard convolution, pooling, and fully-connected layers) 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. We also outperform existing meta-modeling approaches for network design on image classification tasks.