A Particle Swarm Optimization-based Flexible Convolutional Auto-Encoder for Image Classification
This addresses the problem of limited neural network architecture design for image classification researchers, though it appears incremental as it builds on existing auto-encoder and optimization techniques.
The authors tackled the problem of convolutional auto-encoders being unable to construct state-of-the-art convolutional neural networks due to architectural constraints, by proposing a flexible convolutional auto-encoder with an architecture discovery method using particle swarm optimization. Experimental results on four image classification datasets showed that their work significantly outperformed peer competitors including the state-of-the-art algorithm.
Convolutional auto-encoders have shown their remarkable performance in stacking to deep convolutional neural networks for classifying image data during past several years. However, they are unable to construct the state-of-the-art convolutional neural networks due to their intrinsic architectures. In this regard, we propose a flexible convolutional auto-encoder by eliminating the constraints on the numbers of convolutional layers and pooling layers from the traditional convolutional auto-encoder. We also design an architecture discovery method by using particle swarm optimization, which is capable of automatically searching for the optimal architectures of the proposed flexible convolutional auto-encoder with much less computational resource and without any manual intervention. We use the designed architecture optimization algorithm to test the proposed flexible convolutional auto-encoder through utilizing one graphic processing unit card on four extensively used image classification datasets. Experimental results show that our work in this paper significantly outperform the peer competitors including the state-of-the-art algorithm.