Generative Teaching Networks: Accelerating Neural Architecture Search by Learning to Generate Synthetic Training Data

Felipe Petroski Such, Aditya Rawal, Joel Lehman, Kenneth Stanley, Jeff Clune

Keywords: Generative models, generating synthetic data, neural architecture search, learning to teach, meta-learning

TL;DR: We meta-learn a DNN to generate synthetic training data that rapidly teaches a learning DNN a target task, speeding up neural architecture search nine-fold.

Abstract: This paper investigates the intriguing question of whether we can create learning algorithms that automatically generate training data, learning environments, and curricula in order to help AI agents rapidly learn. We show that such algorithms are possible via Generative Teaching Networks (GTNs), a general approach that is applicable to supervised, unsupervised, and reinforcement learning. GTNs are deep neural networks that generate data and/or training environments that a learner (e.g.\ a freshly initialized neural network) trains on before being tested on a target task. We then differentiate \emph{through the entire learning process} via meta-gradients to update the GTN parameters to improve performance on the target task. GTNs have the beneficial property that they can theoretically generate any type of data or training environment, making their potential impact large. This paper introduces GTNs, discusses their potential, and showcases that they can substantially accelerate learning. We also demonstrate a practical and exciting application of GTNs: accelerating the evaluation of candidate architectures for neural architecture search (NAS), which is rate-limited by such evaluations, enabling massive speed-ups in NAS. GTN-NAS improves the NAS state of the art, finding higher performing architectures when controlling for the search proposal mechanism. GTN-NAS also is competitive with the overall state of the art approaches, which achieve top performance while using orders of magnitude less computation than typical NAS methods. Overall, GTNs represent a first step toward the ambitious goal of algorithms that generate their own training data and, in doing so, open a variety of interesting new research questions and directions.