System and method for parameter compression of capsule networks using deep features

A system comprising:

• one or more memories;
• and
• one or more first hardware processors, the one or more first memories coupled to the one or more first hardware processors, wherein the one or more first hardware processors are configured to execute programmed instructions stored in the one or more first memories, to:
• employ a deep feature based Capsule Network for a task, the employed deep feature based Capsule Network is a deep neural network including one or more capsules, each capsule of the one or more capsules is a set of neurons of the deep feature based Capsule Network, the deep feature based Capsule Network comprising a feature extraction layer comprising a set of feature extraction units, wherein each feature extraction unit of the set of feature extraction units comprises alternate convolutional layers with 512 filters having kernal size 3×3 and 256 filters with kernal size 1×1 respectively, and a batch normalization (BN) layer, a capsule layer that includes a primary capsule layer and a class capsule layer, the primary capsule layer comprising a set of primary Capsules to define representation of initial deep features in form of vectors that are trained to learn geometric transformations in a routing-by-agreement process, and the class capsule layer comprising a set of class capsules, the feature extraction layer is prior to the capsule layer, the feature extraction layer configured to extract the deep features from input data,
• wherein the primary capsule layer comprises a plurality of neurons of the deep feature based Capsule Network,
• wherein the routing-by-agreement process comprises passing information from each of the set of primary capsules to each capsule in the set of class capsules only for features that contributed to prediction of at least one class in the past, and refrains from passing the information from each of the set of primary capsules to each capsule in the set of class capsules for features that did not contribute to prediction of at least one class in the past,
• wherein to employ the deep feature based Capsule Network for the task, the one or more first hardware processors are configured by the instructions to:
• pass, successively through each of the set of feature extraction units, an input data to obtain set of deep features;
• perform, by the primary capsule layer, a convolution of the set of deep features into the set of primary capsules, wherein the set of primary capsules in the primary capsule layer comprises an optimal number of capsules based on a number of features in the set of deep features;
• and
• predict, at the class capsule layer, a set of classes from the set of deep features, wherein each capsule in the set of class capsules predict a single class and is activated only for those primary block capsules which together agrees for a required class during a training phase of the deep feature based Capsule Network, wherein a number of capsules of the one or more capsules in the set of class capsules is equal to a number of classes associated with the task, and wherein the deep feature based Capsule Network further comprises a decoder communicatively coupled with the class capsule layer, the decoder capable of receiving the set of classes predicted by the class capsule layer and reconstructing an image associated with the task based on the set of classes during a training phase of the deep feature based Capsule Network.

One or more non-transitory machine readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause managing a plurality of events, the instructions cause:

• employing a deep feature based Capsule Network for a task, the employed deep feature based Capsule Network is a deep neural network including one or more capsules, each capsule of the one or more capsules is a set of neurons of the deep feature based Capsule Network, the deep feature based Capsule Network comprising:
• a feature extraction layer comprising a set of feature extraction units wherein each feature extraction unit of the set of feature extraction units comprises alternate convolutional layers with 512 filters having kernal size 3×3 and 256 filters with kernal size 1×1 respectively, and a batch normalization (BN) layer, and
• a capsule layer including:
• a primary capsule layer and a class capsule layer, the primary capsule layer comprising a set of primary Capsules for defining representation of initial deep features in form of vectors that are trained to learn geometric transformations in a routing-by-agreement process, wherein the primary capsule layer comprises a plurality of neurons of the deep feature based Capsule Network, and the class capsule layer comprising a set of class capsules, wherein the routing-by-agreement process comprises passing information from each of the set of primary capsules to each capsule in the set of class capsules only for features that contributed to prediction of at least one class in the past, and refrains from passing the information from each of the set of primary capsules to each capsule in the set of class capsules for features that did not contribute to prediction of at least one class in the past,
• wherein the feature extraction layer is prior to the capsule layer, the feature extraction layer configured to extract the deep features from input data,
• wherein employing the deep feature based Capsule Network comprises:
• passing, successively through each of the set of feature extraction units, an input data to obtain a set of deep features;
• performing, by the primary capsule layer, a convolution of the set of deep features into the set of primary capsules, wherein the set of primary capsules in the primary capsule layer comprises an optimal number of capsules based on a number of features in the set of deep features;
• and
• predicting, at the class capsule layer, a set of classes from the set of deep features, wherein each capsule in the set of class capsules predict a single class and is activated only for those primary block capsules which together agrees for a required class during a training phase of the deep feature based Capsule Network, wherein a number of capsules of the one or more capsules in the set of class capsules is equal to a number of classes associated with the task, and further receiving the set of classes predicted by the class capsule layer via a decoder and reconstructing an image associated with the task based on the set of classes during a training phase of the deep feature based Capsule Network.