Which, when, and how should I use each function, e.g., Dense, Conv2D, BatchNormalization, …, and the number of neurons and filters and layers?
I know that depending on some values that I guess sometimes overfitting occurs, sometimes it predicts very poorly, sometimes it predicts correctly.
What logic should I follow?
What should I include in a sequence for prediction?
Is there a book or something similar that teaches the best practices for building a good model?
tf.keras.Sequential([...])
tiago is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.
To build a Sequential model in TensorFlow, use tf.keras.Sequential(). Add layers like Dense, Conv2D, or LSTM to create a simple stack of layers for your neural network.