Branching and Merging Convolutional Network with Capsules and Residual Connections

A modified implementation of " A Branching and Merging Convolutional Network with Homogeneous Filter Capsules" paper

Changes made from the original architecture are:-

Using original Capsule network instead of Homogeneous Filter Capsules
Using Residual Connections after every two convolution layer
added a trainable coefficient which determines the strength of residual connection, it is similiar to the idea of highboost filtering

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
.ipynb_checkpoints		.ipynb_checkpoints
.gitattributes		.gitattributes
Branching_and_Merging_Residual_CapsNet_CIFAR10.ipynb		Branching_and_Merging_Residual_CapsNet_CIFAR10.ipynb
Branching_and_Merging_Residual_CapsNet_FashionMNIST.ipynb		Branching_and_Merging_Residual_CapsNet_FashionMNIST.ipynb
Branching_and_Merging_Residual_CapsNet_MNIST.ipynb		Branching_and_Merging_Residual_CapsNet_MNIST.ipynb
README.md		README.md
Testing Modified Model CIFAR10.ipynb		Testing Modified Model CIFAR10.ipynb
Testing Modified Model FashionMNIST.ipynb		Testing Modified Model FashionMNIST.ipynb
Testing Modified Model MNIST.ipynb		Testing Modified Model MNIST.ipynb

Provide feedback