Principal Component Analysis on Convolutional Neural Network Using Transfer Learning Method for Image Classification of Cifar-10 Dataset

The current era was defined by an overwhelming abundance of information, including multimedia data such as audio, images, and videos. However, with such an enormous amount of image data available, accurately and efficiently selecting the necessary images poses a significant challenge. To address this, image classification has emerged as a viable solution for organizing and managing large volumes of image data, thereby mitigating the issue of cluttered image datasets. One of the most popular algorithms for image classification is the Convolutional Neural Network (CNN), which reduces the complexity of network structure and parameters by leveraging local receptive fields, weight sharing, and pooling operations. CNN is a type of artificial neural network specifically designed to process grid-like data, such as images, using convolutional layers to automatically detect local features. Nonetheless, CNN faces several challenges, such as gradient diffusion, large dataset requirements, and slow training processes. To overcome these issues, Transfer Learning has been widely adopted in CNN-based image classification, and Principal Component Analysis (PCA) has been employed to accelerate the training process. PCA is a technique used to reduce data dimensionality by identifying the principal components that account for most of the variance in the data. This study tested the efficacy of PCA-based CNN architecture using the Transfer Learning method on the Cifar-10 dataset. The results demonstrated that the PCA-based CNN architecture achieved the highest accuracy, with a testing accuracy rate of 0.8982 (89%).