Learning Rate Optimization in CNN for Accurate Ophthalmic Classification

One of the most important hyper-parameters for model training and generalization is learning rate. Recently, many research studies have shown that optimizing rate schedule very useful deep neural networks to get accurate efficient results. In this paper, different schedules using some comprehensive optimization techniques been compared in order measure accuracy a convolutional network CNN classify four ophthalmic conditions. work, based on Keras TensorFlow has deployed Python database contains 1692 images, which consists types cases: Glaucoma, Myopia, Diabetic retinopathy, Normal eyes. The trained Google Colab. GPU with adaptive algorithms. Constant rate, time-based decay, step-based exponential addressed. Adam method. outperformed other achieved best 92.58% set 80.49% validation datasets, respectively.