FSC-CapsNet: Fractionally-Strided Convolutional Capsule Network for Complex Data.

Jian-wei Liu,Feng Gao,Run-kun Lu,Yuan-feng Lian,Dian-zhong Wang,Xiong-lin Luo,Chu-ran Wang

DOI: https://doi.org/10.1109/ijcnn.2019.8851924

2019-01-01

Abstract:Recently, a novel neural network called CapsNet has attracted the attention of many researchers. It is a great attempt to overcome the drawback of convolutional neural networks (CNNs) and achieves state-of-the-art performance on some simple datasets like MNIST. However, this network architecture is built specifically for MNIST and gets a poor performance on more complex datasets like CIFAR-10. To address this problem, aiming at complex data, we propose a new CapsNet architecture called Fractionally-Strided Convolutional Capsule Network (FSC-CapsNet). We modify both network structures of the encoder and decoder of CapsNet. For the purpose of extracting better features, we increase the number of convolutional layers before capsule layer in the encoder and improve the reconstruction performance by adopting two fractionally-strided convolutional layers in the decoder. In addition, no pooling layers are used in our architecture. To assess the performance of our proposed network on complex data, we conduct experiments with a single model without using any ensembled methods and data augmentation techniques on five real-world datasets, which are of higher dimensionality and larger size than MNIST. The experimental results demonstrate that our proposed method achieves better performance and improves the reconstruction performance compared with the normal CapsNet.