What does the solution with size 1 mean? For each run of function prediction, it yields a number. My dataset has 105 entries and the test and validation sets are in a 70:30 ratio. The target expression has a single input variable.
The size is always the sum of the complexities of the building blocks that appear in a solution. Most often the simplest solution (size of 1) will be just a constant - typically the mean of the target variable.
Is there a method for the solution to not include constant offset added to the remaining functional solution? I have a frequency study of a bracket. The solution from Eureqa continues to find solutions that start with a constant offset. Frequencies do not have that behavior, although real problems do not go to 0 Hz.
The solution does need to have constants in the solution, e.g. multiplying the function by 1/pi for example.
One way to do this is to use the target expression in the advanced settings to force the models to multiply all terms by a variable. So if you know a variable you want to force into the model, you can do something like: