15.15 Loss functions and optimizers
- Once network architecture (
model
) is defined → choose loss function & optimizer - Loss function (also called objective function)
- Quantity that will be minimized during training (measure of success)
- Optimizer
- Determines how the network will be updated based on the loss function
# Example code: compile the model
%>% compile(
model optimizer = "rmsprop", # Optimizer algorithm
loss = "mse", # Mean squared error loss function
metrics = c("mae") # mean absolute error (MAE)
)