Two Classes Classification Using Different Optimizers in Convolutional Neural Network
Ahmad Shaf, Tariq M. Ali, Wajiha Farooq, Sarah Javaid, Umar Draz, Sana Yasin · 2018
Convolutional Neural Network (CNN) is the most powerful architecture of pattern recognition problems in machine learning. CNN have ability to automatically detect relevant data from training examples. In CNN, different optimizers are used for the classification purposes. Each optimizer performs well in its predefined problem and parameters. In this paper, we have taken the dataset of cats and dogs for classification purposes and evaluated the performance of multiple optimizers against our designed CNN model with different learning rate and epochs. We have tested it on CPU and computed optimizers performance in term of accuracy and error rate. The results shows that after 5 epochs, momentum optimizer performs well as compared to others for the classification problem of two classes.