Hybrid Convolutional Neural Network In Complex Domain And Its Applications

Abstract

A convolutional neural network (CNN or ConvNet) is a category of artificial neural network which is commonly used to interpret hierarchical patterns of visual data including video and images. A special class of neural network named deep CNN has been found to be effective in many applications in the field of image processing and computer vision. Image classification, segmentation, object detection, facial recognition, video processing and natural language processing are some of the most stimulating application areas of CNN which are currently in vogue. Presently, the usage of deep learning concepts for the classification of medical images has been in the limelight of many researchers as it has outperformed the other established and standardized methods. Many fascinating architectures pertaining to deep CNN have been developed and recommended for various applications, especially in computer vision and the healthcare domain. Existing machine learning models for object recognition and pattern classification are found to be obsolete as they use handcrafted feature extraction techniques which are time-consuming and remain fail to include all the relevant features. The robustness of the overall pattern recognition of such models may not be consistent with data translation variance, hierarchical feature representation, parameter sharing, local variation, dynamic scalability, etc. Such an Existing model can easily achieve local optima for pattern identification but remain inconsistent to formulate global optima due to the unprecedented variations in the input dataset. newlineA novel architecture named Hybrid Convolutional Neural Network in Complex Domain (HCNN-CD) has been proposed. The purpose of this study is to utilize a model based on complex-valued hybrid CNN that claims to have resolved major challenges and help to develop optimized applications based on images. Hybrid CNN incorporates the architecture that combines two or more networks of convolutional layers. The model is able to establish a relationship between spatia