如何在PyGAD中提供命名参数

Question

我看过关于用PyGAD优化任意函数的教程。作为一个简单的POC，我想实现线性回归。我对与Keras或PyTorch集成不感兴趣；我知道集成是受支持的，但这比必要的火力更大。

这是本教程的简单说明函数，它在权重上收敛，使得X[i] . W[i] = value:

def fitness_func(solution, solution_idx):
output = numpy.sum(solution*function_inputs)
fitness = 1.0 / numpy.abs(output - desired_output)
return fitness

以下是我对线性回归的尝试：

#  X,         Y
# [[1,3,5,7], [2,4,6,8]]
# slope = 1
# intercept = 1 
def fitness_function_factory(slope=0, intercept=0, function_inputs = numpy.array([[1,3,5,7], [2,4,6,8]]), desired_output = 0 ):
def fitness_function(solution, solution_idx):
pred = slope * function_inputs[0] + intercept
error = numpy.sum(numpy.abs(pred - function_inputs[1]))
fitness = 1/(error - desired_output)
return fitness 
return fitness_function
lr_ga = pygad.GA(num_generations=num_generations,
num_parents_mating=num_parents_mating,
fitness_func=fitness_function_factory(0,0),
sol_per_pop=sol_per_pop,
num_genes=num_genes,
init_range_low=init_range_low,
init_range_high=init_range_high,
parent_selection_type=parent_selection_type,
keep_parents=keep_parents,
crossover_type=crossover_type,
mutation_type=mutation_type,
mutation_percent_genes=mutation_percent_genes)
lr_ga.run()

以及相应的错误：

7 def linear_regression(solution, solution_idx):
----> 8   pred = solution['slope'] * function_inputs[0] + solution['intercept']
9   error = numpy.abs(pred - function_inputs[1])
10   fitness = 1/(error - desired_output)
IndexError: only integers, slices (`:`), ellipsis (`...`), numpy.newaxis (`None`) and integer or boolean arrays are valid indices

注：我从这个问答中获得了灵感，使用了一个工厂函数来包装健身函数。

我应该如何启用命名参数，以便可以检索斜率和截距？

Answer 1

num_generations = 50
num_parents_mating = 4
sol_per_pop = 8
num_parameters = 2 
init_range_low = -2
init_range_high = 5
parent_selection_type = "sss"
keep_parents = 1
crossover_type = "single_point"
mutation_type = "random"
mutation_percent_genes = 10
def fitness_function_factory(X=numpy.array([1,3,5,7]), Y=numpy.array([2,4,6,8]), desired_output = 0 ):
def fitness_function(solution, solution_idx):
pred = solution[0] * X + solution[1]
errors = numpy.abs(pred - Y)
fitness = 1/(numpy.sum(errors) - desired_output)
return fitness 
return fitness_function
lr_ga = pygad.GA(num_generations=100,
num_parents_mating=num_parents_mating,
fitness_func=fitness_function_factory(),
sol_per_pop=sol_per_pop,
num_genes=num_parameters,
init_range_low=init_range_low,
init_range_high=init_range_high,
parent_selection_type=parent_selection_type,
keep_parents=keep_parents,
crossover_type=crossover_type,
mutation_type=mutation_type,
mutation_percent_genes=mutation_percent_genes)
lr_ga.run()
lr_ga.best_solution()[0]
>>>
array([0.98365965, 1.02723423]) # slope, intercept