I would like to use DEAP's genetic programming functionality to randomly generate some short Python scripts (that I'll later want to mutate and evolve). While learning how to use it, I am encountering an exception that I do not understand.
The following is a minimal program that recreates the exception. It is intended to ask DEAP to generate random programs that add a small random positive integer to a random number from a Gaussian distribution. In this minimal example the input to the program is ignored.
import numpy as np
from deap import base, creator, gp, tools
import operator
class POSINT() : pass
class GAUSS() : pass
pset = gp.PrimitiveSetTyped('main',[str],float)
pset.addPrimitive(operator.add, [POSINT,GAUSS],float)
pset.addEphemeralConstant('POSINT',lambda: np.random.randint(10),POSINT)
pset.addEphemeralConstant('GAUSS',lambda: np.random.randn(),GAUSS)
creator.create("Individual", gp.PrimitiveTree)
toolbox = base.Toolbox()
toolbox.register("expr", gp.genHalfAndHalf, pset=pset, min_=1, max_=2)
toolbox.register("individual", tools.initIterate, creator.Individual, toolbox.expr)
expr = toolbox.individual()
print(expr)
Code explanation: To constrain the inputs to the allowed ranges (positive integers and Gaussian), I define two types (T_POSINT and T_GAUSS). I then add a single primitive operation (operator.add) which takes two arguments (one of each allowed type) and returns a float.
Using the same code as the tutorial linked above, I try to generate an 'individual' (i.e. an instance of a program that satifies the constraints I have specified. This program works if the min_ and max_ arguments are both set to 1, but if I increase max to 2, the program raises the following exception. Why? I haven't forced it to make a tree of size 2, just increased the maximum accepted size.
Traceback (most recent call last):
File "/home/user/.local/lib/python3.8/site-packages/deap/gp.py", line 624, in generate
prim = random.choice(pset.primitives[type_])
File "/usr/lib/python3.8/random.py", line 290, in choice
raise IndexError('Cannot choose from an empty sequence') from None
IndexError: Cannot choose from an empty sequence
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "stack_overflow_deap.py", line 19, in <module>
expr = toolbox.individual()
File "/home/user/.local/lib/python3.8/site-packages/deap/tools/init.py", line 52, in initIterate
return container(generator())
File "/home/user/.local/lib/python3.8/site-packages/deap/gp.py", line 574, in genHalfAndHalf
return method(pset, min_, max_, type_)
File "/home/user/.local/lib/python3.8/site-packages/deap/gp.py", line 557, in genGrow
return generate(pset, min_, max_, condition, type_)
File "/home/user/.local/lib/python3.8/site-packages/deap/gp.py", line 627, in generate
raise IndexError("The gp.generate function tried to add " \
File "/home/user/.local/lib/python3.8/site-packages/deap/gp.py", line 624, in generate
prim = random.choice(pset.primitives[type_])
File "/usr/lib/python3.8/random.py", line 290, in choice
raise IndexError('Cannot choose from an empty sequence') from None
IndexError: The gp.generate function tried to add a primitive of type '<class '__main__.POSINT'>', but there is none available.
Compilation exited abnormally with code 1 at Wed May 26 12:35:38
Why is it saying The gp.generate function tried to add a primitive of type <class '__main__.POSINT'>, but there is none available when there is clearly a terminal of that type (the POSINT ephemeral constant)?
More generally, how should I avoid this exception? In this trivial case, the solution of setting max_=1 works, but in the more complex cases I'd like to be able to generate different size trees. Should I just repeat the expr = toolbox.individual() command until it doesn't return an exception? Seems a bit hacky--no? Thanks for your time!