An Efficient Approach for Using Expectation Maximization Algorithm in Capsule Networks
Paper ID : 1019-MVIP2020
Moein Hasani *1, Amin Nasim Saravi2, Hassan Khotanlou2
2Department of Computer Engineering Bu Ali Sina University
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 dynamic routing between capsules in "Sabour et al" and in a more recent paper "Matrix Capsules with EM Routing" they proposed a more complete version with Expectation-Maximization (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-the-art results on the smallNORB dataset and it 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 for EM routing iterations. For our research, we use three datasets: Yale face dataset, Belgium TS (traffic sign) dataset, and FashionMNIST (FMNIST) dataset. Unlike the past studies, we have tried using un-equal numbers of EM routing iterations for different stages of the CapsNet. Our research proposes efficient settings for EM routing iterations of CapsNets.
Capsule Networks, Routing-by-Agreement, Convolutional Neural Networks, CNNs
Status : Paper Accepted (Poster Presentation)