Fixing stop_fitness for QueensOpt with custom fitness fn

[eko-wibowo]

Fixing the issue described here by allowing custom stop_fitness instead of using the default for custom fitness function

#32

Repro problem

def get_queens_problem(N) -> mlrose.QueensOpt:
    def queens_max(state):
        # Initialize counter
        fitness = 0

        # For all pairs of queens
        for i in range(len(state) - 1):
            for j in range(i + 1, len(state)):

                # Check for horizontal, diagonal-up and diagonal-down attacks
                if (state[j] != state[i]) and (state[j] != state[i] + (j - i)) and (state[j] != state[i] - (j - i)):
                    # If no attacks, then increment counter
                    fitness += 1

        return fitness

    fitness_fn = mlrose.CustomFitness(queens_max, problem_type="discrete")
    return mlrose.QueensOpt(length=N, maximize=True, fitness_fn=fitness_fn, stop_fitness=N * (N - 1) / 2)

calls RHCRunner with
iteration_list=np.arange(0, 1001, 100),
restart_list=[10, 25, 100],
max_attempts=500

The plots below show restart as a drop in the fitness function, 3 values in the restarts list are overlapping

• Without stop iter param (never hit the convergence point, took ~20mins). I think some good "path" might have been stopped due to it's hitting fitness 24 (3 * (10-2))
  

• With stop iteration param Hit convergence point (took 2 mins)