What levels should a tensorflow-model have?

Question

I am just learning the TensorFlow and try to write the model for the hate speech detecting in the dataset. So maybe someone have some advices for me, how many and which levels do I need in the model and on what it depends?

Answer 1

This is something that is, for some reason, never taught in deep learning courses. I guess it depends on multiple factors:

1. Which architecture are you going to use? CNN, LSTM, something different? They achieve different goals, they use different layers, and they can also be combined.

2. How much time to you want to spend training the model? Typically, more layers will increase the accuracy, but increase the training time and required epochs a lot.

3. How much data do you have? I'm not 100% sure here, but the larger your model is, the more data it needs to learn well.

If you're just starting, I recommend to focus on one type of neural network and experiment with different layers to see what effects they bring.