부산대학교 도서관

Most common machine learning algorithms were designed to tackle particular problems in seclusion. Based on the business requirements, a machine learning algorithm was employed for a particular task and these algorithms were applied to train models separately on specific feature sets. Most of the machine learning models are trained on the assumption that the test and training sets have similar features while sharing the same underlying distributions. Convolutional Neural Networks (CNNs) have demonstrated remarkable potential for breast cancer detection using histopathology images. Moving forward with the latest research, we present VGG-16 architecture for breast cancer detection, which involves the use of pooling and convolutional layers to extract high-level accuracy from histopathology images. The proposed network consists of sixteen layers, each of which performs convolutional operations on the image to detect local features and pooling operations with the aim of simplifying the feature maps’ dimensionality. The test results illustrate that the VGG-16 architecture secures superior classification accuracy when compared with the previous cutting-edge models.