我想使用import java.util.Arrays;
import java.util.Collections;
import java.util.Comparator;
import java.util.List;
public class Main {
public static void main(String[] args) {
List<String> list = Arrays.asList("Rec1 1 1", "Rec2 1 5", "Rec3 1 3", "Rec4 2 6", "Rec5 2 10)");
Collections.sort(list, new Comparator<String>() {
@Override
public int compare(String one, String two) {
return Integer.valueOf(one.split(" ")[1]).compareTo(Integer.valueOf(two.split(" ")[1]));
}
});
System.out.println(list);
}
}
来最小化损失函数。因为我的代码中的SciPy.optimize.fmin_l_bfgs_b旨在一次运行返回loss_calculator和loss并执行一次评估是昂贵的，所以我必须以这种方式“欺骗”优化器：
loss_prime这只是一个玩具代码，用以证明我的想法。如您所见，方法import numpy as np
from scipy.optimize import fmin_l_bfgs_b as optimizer
x_true = np.arange(0,10,0.1)
m_true = 2.5
b_true = 1.0
y_true = m_true*x_true + b_true
class test:
def __init__(self, x_true, y_true):
self.x_true = x_true
self.y_true = y_true
def func(self, params):
x = self.x_true
y = self.y_true
m, b = params
y_model = m*x+b
error = y-y_model
self.loss = sum(error**2)
temp_dm = sum(2*x*(m*x+b-y))
temp_db = sum(2*(m*x+b-y))
self.dloss = np.asarray([temp_dm, temp_db])
def get_loss(self, params):
self.params = params
self.func(params)
return self.loss
def get_dloss(self, params):
assert (self.params == params).all()
return self.dloss
temp_instance = test(x_true, y_true)
initial_values = np.array([1.0, 0.0])
x = optimizer(temp_instance.get_loss, initial_values, temp_instance.get_dloss)
实际上什么也没做，get_dloss是在我们调用loss_prime时计算出来的。
此玩具代码适用于此玩具示例，但我不确定这在实际实现中是否合法。
我担心的是get_loss首先会评估SciPy.optimize.fmin_l_bfgs_b吗？