PyPI软件包的“密度”密度比估计分数均为nan

问题描述

我正在使用densratio Density ratio estimation Python库来获取训练数据的KIELP权重。我的数据集是 kaggle 的房价，经过一轮热编码后，共有222列。

x_train.shape：

(978,222)

x_test.shape：

(482,222)

调用后，所有分数均为nan。

from densratio import densratio

densratio(x_train,x_test,alpha=0.1,sigma_range=[0.1,0.3,0.5,0.7,1],lambda_range=[0.01,0.02,0.03,0.04,0.05]))

RuLSIF starting...
Searching for the optimal sigma and lambda...
sigma = 0.10000,lambda = 0.01000,score = nan
sigma = 0.10000,lambda = 0.02000,lambda = 0.03000,score = nan

--------- ....... --------- ----------------all the same between


sigma = 1.00000,score = nan
sigma = 1.00000,lambda = 0.04000,lambda = 0.05000,score = nan
Found optimal sigma = 0.000,lambda = 0.000.
Optimizing theta...
Approximate alpha-relative PE-divergence = nan
Approximate alpha-relative KL-divergence = nan
RuLSIF completed.
Method: RuLSIF

Alpha: 0.1

Kernel information:
  Kernel type: Gaussian
  Number of kernels: 100
  Bandwidth(sigma): 0
  Centers: array([[2.000e+00,4.000e+00,2.000e+00,...,2.022e+03,..

Kernel Weights (theta):
  array([nan,nan,..

Regularization Parameter (lambda): 0

Alpha-Relative PE-divergence: nan

Alpha-Relative KL-divergence: nan

Function to Estimate Density Ratio:
  compute_density_ratio(x)

解决方法

暂无找到可以解决该程序问题的有效方法，小编努力寻找整理中！

如果你已经找到好的解决方法，欢迎将解决方案带上本链接一起发送给小编。

小编邮箱:dio#foxmail.com (将#修改为@）

kernel-density python