Deep Reinforcement Learning for Image Classification

We present a deep reinforcement learning-based approach that leverages data augmentation techniques to improve the accuracy of image classification tasks. We define the essential components of deep reinforcement learning and create an environment that enables the model to learn more autonomously when additional learning is required. Our experiments with various benchmark datasets(CIFAR-10, SVHN and WM-811K) demonstrate that the proposed deep reinforcement learning-based approach outperforms the conventional convolutional neural networks in terms of classification accuracy, highlighting its potential to effectively address classification problems using deep reinforcement learning.