Distributed Evolution of Deep Autoencoders
This work addresses the problem of automating autoencoder design for researchers and practitioners in machine learning, though it appears incremental as it builds on existing evolutionary methods.
The authors tackled the challenge of designing autoencoders for specific tasks by developing a distributed system that uses an evolutionary algorithm to optimize modular autoencoder architectures. The system outperformed random search by nearly an order of magnitude on manifold learning and image denoising tasks and achieved near-linear scaling with additional worker nodes.
Autoencoders have seen wide success in domains ranging from feature selection to information retrieval. Despite this success, designing an autoencoder for a given task remains a challenging undertaking due to the lack of firm intuition on how the backing neural network architectures of the encoder and decoder impact the overall performance of the autoencoder. In this work we present a distributed system that uses an efficient evolutionary algorithm to design a modular autoencoder. We demonstrate the effectiveness of this system on the tasks of manifold learning and image denoising. The system beats random search by nearly an order of magnitude on both tasks while achieving near linear horizontal scaling as additional worker nodes are added to the system.