Efficacy of Deep Learning Using Softmax Regression in Comparison with the Single Layer to Multi-Layer Convolution Network in Predicting Cervical Cancer

Peter Kulandai Raj and Dr.S. Sheeja


Biologically inspired programming paradigm isan �Artificial Neural network� (ANN). This has been in the realm of research since 1940�s. An idea that employs the fundamental technique of brain function in processing the information and taking decision instantaneously. The objective of automation is to simplify the laborious tasks where human intervention is paramount. Starting from late 1950�s people started to think of employing computers at a larger extent, and thus computers started dominating from the automation perspective. Instructions are fed to computer and it is operating based on the instructions. In course of time this sounds very interesting however Machine Learning raises the bar of smart computing to another level; making the computer to infer the decisions just by looking at the patterns of sample data with facts and decide new solutions instantaneously on the given problem with new constraints. Especially the Advent of Deep learning in Neural Network (DNN) has transformed the way of machine learning has evolved to provide more accurate solutions. Neural Network (NN) although is not a new concept, however last couple of years DNN has started dominating the machine learning community. Machine Learning competitions like Kaggle has given the opportunity to prove the importance of deep Learning to solve the complex problems with more accuracy. This paper sheds an insight into Cervical Cancer Prediction and correlates how deep learning can predict cervical cancer with higher accuracy of 98.9%. Also, discussing employing the �softmax regression� technique with single and multi-layer Convolution Neural Networks (CNN) to improve the performance of 91.9% using single layer NN to obtaining 98.9% accuracy in the multilayer NN for pap smear images.

Issue: 01-Special Issue

Year: 2017

Pages: 139-148

Purchase this Article