EvoAE -- A New Evolutionary Method for Training Autoencoders for Deep Learning Networks

Although deep learning has achieved outstanding performances on several difficult machine learning applications, there are multiple issues that make its application on new problems difficult: speed of training, local minima, and manual selection of hyper-parameters. To overcome these problems, this paper proposes a new evolutionary method, EvoAE, to train auto encoders for deep learning networks. By evolving a population of auto encoders, EvoAE learns multiple features in each auto encoder in the form of hidden nodes, evaluates the auto encoders based on their reconstruction quality, and generates new auto encoders using crossover and mutation with chromosomes made up of hidden nodes and associated connections and weights. EvoAE optimizes network weights and structures of auto encoders simultaneously and employs a mini-batch variant, called Evo-batch, to speed up auto encoder search on large datasets. Furthermore, EvoAE supports different training methods in data partitioning and selection, requires little manual intervention, and reduces overall training time drastically over traditional methods on large datasets.