I am building a simple NN for a regression analysis (0.5 mln rows of data) in Keras.
# Define model model = Sequential() model.add(Dense(32, activation = 'relu')) model.add(Dense(32, activation = 'sigmoid')) model.add(Dense(1)) # Define model parameters model.compile(optimizer = 'sgd', loss = 'mse', metrics = [tf.keras.metrics.MeanAbsoluteError()])
I am satisfied with results, but would like to know if there is any rule of thumb or recommendations how to build such models for a regression analysis.
- How many layers?
- Which activation functions?
- Which optimizer?
What is your state of the art approach, when you start to do such type of analysis?
Especially appreciate any academic sources.