Greetings! In fine-tuning pre-trained models (transfer learning), I’m seeking guidance on selecting the architecture for the final layers. Specifically, I’d appreciate insights on determining the optimal number of dense layers, dropout rates, and neurons.How do you select the final layers you want to train youself.
This is mainly experimentation. However, the overall final layer depends on the task you are working on. If it’s a classification, then you look at the number of classes you want to predict.
1 Like
i am working on brain tumor detection.Can you guide me for what layers should i use.Thanks