- Abstract: In deep learning, performance is strongly affected by the choice of architecture and hyperparameters. While there has been extensive work on automatic hyperpa- rameter optimization for simple spaces, complex spaces such as the space of deep architectures remain largely unexplored. As a result, the choice of architecture is done manually by the human expert through a slow trial and error process guided mainly by intuition. In this paper we describe a framework for automatically designing and training deep models. We propose an extensible and modular lan- guage that allows the human expert to compactly represent complex search spaces over architectures and their hyperparameters. The resulting search spaces are tree- structured and therefore easy to traverse. Models can be automatically compiled to computational graphs once values for all hyperparameters have been chosen. We can leverage the structure of the search space to introduce different model search algorithms, such as random search, Monte Carlo tree search (MCTS), and sequen- tial model-based optimization (SMBO). We present experiments comparing the different algorithms on CIFAR-10 and show that MCTS and SMBO outperform random search. We also present experiments on MNIST, showing that the same search space achieves near state-of-the-art performance with a few samples. These experiments show that our framework can be used effectively for model discov- ery, as it is possible to describe expressive search spaces and discover competitive models without much effort from the human expert. Code for our framework and experiments has been made publicly available
- TL;DR: We describe a modular and composable language for describing expressive search spaces over architectures and simple model search algorithms applied to these search spaces.
- Keywords: architecture search, deep learning, hyperparameter tuning