并行循环GMM / EM群集

Question

我很难找到并行拟合高斯混合EM聚类模型的正确方法。

我有一个带有 12核的Windows，我想利用它。

到目前为止，我已经关注了这些链接。

here

Implement Parallel for loops in Python

我必须遵循我想并行运行的EM集群模型的实现。 如果嵌套循环有问题，我可以让外部循环并行运行。

maxCLust=18
from sklearn import mixture
lowest_bic = np.infty
bic = []
n_components_range = range(1, maxCLust)
cv_types = ['spherical', 'tied', 'diag', 'full']
methods_in=['kmeans', 'random']
for method_in in methods_in:
  for cv_type in cv_types:
      for n_components in n_components_range:
          # Fit a Gaussian mixture with EM
          gmm = mixture.GaussianMixture(n_components=n_components,
                                        covariance_type=cv_type,
                                        init_params=method_in)
          gmm.fit(np.asarray(X))
          print(n_components)
          bic.append(gmm.bic(np.asarray(X)))
          if bic[-1] < lowest_bic:
              lowest_bic = bic[-1]
              best_gmm = gmm

bic = np.array(bic)
bestEMmodel = best_gmm
labels = bestEMmodel.predict(X)
bestEMmodel.n_components
labels

这是我最后一次尝试

import multiprocessing as mp
nprocs = mp.cpu_count()
print(f"Number of CPU cores: {nprocs}")

pool = mp.Pool(processes=nprocs)

from sklearn import mixture

def GMMbic (n_components_range, myData):

  lowest_bic = np.infty
  bic = []
  n_components_range = n_components_range
  cv_types = ['spherical', 'tied', 'diag', 'full']
  methods_in=['kmeans', 'random']
  for method_in in methods_in:

    for cv_type in cv_types:
        for n_components in n_components_range:
            # Fit a Gaussian mixture with EM
            gmm = mixture.GaussianMixture(n_components=n_components,
                                          covariance_type=cv_type,
                                          init_params=method_in)
            gmm.fit(np.asarray(myData))
            bic.append(gmm.bic(np.asarray(myData)))
            if bic[-1] < lowest_bic:
                lowest_bic = bic[-1]
                best_gmm = gmm
                return best_gmm

results = [pool.apply(GMMbic, args=(18,X))]