Capsule Networks caught the world’s attention as a technique said to be a viable replacement for the established practice of Convolutional Neural Networks, providing advantage of viewpoint invariance by keeping the model closer to how human vision works. The work by Sabour Et. Al. introduced a novel method of encapsulating information about state of features in vector form. For this to work, a new vector-in vector-out activation function termed ‘Squash’ was introduced. Selection and parametrization of activation functions in CNNs has a vital role in determination of model accuracy, however due to nascent nature of Capsule Networks, no such alternative exists. This work proposes an alternative approach to represent the Squashing function by embedding feature vector in higher dimensional hyperspace. It also presents a method to effectively parametrize the Squashing function through introduction of two additional hyperparameters.