The 11th Iranian and the first International Conference on Machine Vision and Image Processing

An Efficient Approach for Using Expectation Maximization Algorithm in Capsule Networks

Moein Hasani, Amin Nasim Saravi, Hassan Khotanlou
The 11th Iranian and the first International Conference on Machine Vision and Image Processing (MVIP 2020)

Abstract

Capsule Networks (CapsNets) are brand-new architectures that have shown ground-breaking results in certain areas of Computer Vision (CV). In 2017, Hinton and his team introduced CapsNets with routing-byagreement in “Sabour et al” and in a more recent paper “Matrix Capsules with EM Routing” they proposed a more complete architecture with ExpectationMaximization (EM) algorithm. Unlike the traditional convolutional neural networks (CNNs), this architecture is able to preserve the pose of the objects in the picture. Due to this characteristic, it has been able to beat the previous state-of-theart results on the smallNORB dataset, which includes images with various view points. Also, this new architecture is more robust to white box adversarial attacks. However, CapsNets have two major drawbacks. They can’t perform as well as CNNs on complex datasets and, they need a huge amount of time for training. We try to mitigate these shortcomings by finding optimum settings of EM routing iterations for training CapsNets. Unlike the past studies, we use un-equal numbers of EM routing iterations for different stages of the CapsNet. We manage to achieve higher accuracies than the original CapsNet while training the network up to three times faster. For our research, we use three datasets: Yale face dataset, Belgium Traffic Sign dataset, and FashionMNIST dataset

Keywords: Capsule Networks, Routing-by-Agreement, Convolutional Neural Networks, CNNs



© 2017-2021 ISMVIP All Rights Reserved