input/output/recurrent dropout layers in BiLSTM_Classifier and how they affect the model and prediction

Question

I would like to have some understanding/information on the input/output/recurrent dropout layers in BiLSTM_Classifier and how they affect the model and prediction.