Machine learning-based classification of eye problems is a promising topic for the quick and accurate diagnosis of eye diseases. It must also address problems including the dearth of excellent annotated data, the complexity of ocular anatomy, and the constrained availability of data. Machine learning algorithms may introduce bias into the diagnostic process and may not take important non-visual factors like patient history and lifestyle into consideration. This study assesses the effectiveness of several optimizers in categorizing eye illnesses using a deep learning strategy with a ResNet Convolutional Neural Network (CNN) paired with a Bidirectional Long Short-Term Memory (LSTM) network. There are Four different optimizers were tested and their performance was compared in terms of accuracy and loss values: Adagrad, FLTR, NADAM, and SGD. The results show that Adagrad outperformed the other optimizers in terms of accuracy, reaching an average of 95.3% compared to Nadam (79%), FLTR (62%), and SGD (80%). However, all four optimizers produced relatively low loss values, indicating a high level of model convergence and stability. These findings provide valuable insights into the selection of optimizers in the context of eye disease classification using deep learning algorithms.