titu1994/neural-architecture-search
A Keras/TensorFlow implementation of Neural Architecture Search with a Controller RNN that discovers optimal network architectures via reinforcement learning.
Not currently ranked — collecting fresh signals.
star history
This repository implements the Controller RNN from the NAS paper and NASNet paper, using an RL agent to search the space of neural network architectures. The Controller samples actions defining network parameters (kernel size, filters), while a NetworkManager trains generated Keras child networks and computes rewards. The implementation uses StateSpace to define the architecture search space and manages the training loop between controller and child networks.
Frequently asked
- What is titu1994/neural-architecture-search?
- A Keras/TensorFlow implementation of Neural Architecture Search with a Controller RNN that discovers optimal network architectures via reinforcement learning.
- Is neural-architecture-search open source?
- Yes — titu1994/neural-architecture-search is open source, released under the MIT license.
- What language is neural-architecture-search written in?
- titu1994/neural-architecture-search is primarily written in Python.
- How popular is neural-architecture-search?
- titu1994/neural-architecture-search has 446 stars on GitHub.
- Where can I find neural-architecture-search?
- titu1994/neural-architecture-search is on GitHub at https://github.com/titu1994/neural-architecture-search.