Python EM算法的缓冲区溢出
我正在尝试实现高斯混合的EM算法。我以以下教程为基础: 该算法在某些值上运行良好,但当数值较大时,我通常会得到矩阵的无穷大 这是我的代码:Python EM算法的缓冲区溢出,python,algorithm,Python,Algorithm,我正在尝试实现高斯混合的EM算法。我以以下教程为基础: 该算法在某些值上运行良好,但当数值较大时,我通常会得到矩阵的无穷大 这是我的代码: from scipy.stats import multivariate_normal as mvn import numpy as np import matplotlib.pyplot as plt def em_gmm_orig(xs, pis, mus, sigmas, tol=0.01, max_iter=100): n, p = x
from scipy.stats import multivariate_normal as mvn
import numpy as np
import matplotlib.pyplot as plt
def em_gmm_orig(xs, pis, mus, sigmas, tol=0.01, max_iter=100):
n, p = xs.shape
k = len(pis)
ll_old = 0
for i in range(max_iter):
exp_A = []
exp_B = []
ll_new = 0
# E-step
ws = np.zeros((k, n))
for j in range(len(mus)):
for i in range(n):
ws[j, i] = pis[j] * mvn(mus[j], sigmas[j]).pdf(xs[i])
ws /= ws.sum(0)
# M-step
pis = np.zeros(k)
for j in range(len(mus)):
for i in range(n):
pis[j] += ws[j, i]
pis /= n
mus = np.zeros((k, p))
for j in range(k):
for i in range(n):
mus[j] += ws[j, i] * xs[i]
mus[j] /= ws[j, :].sum()
sigmas = np.zeros((k, p, p))
for j in range(k):
for i in range(n):
ys = np.reshape(xs[i]- mus[j], (2,1))
sigmas[j] += ws[j, i] * np.dot(ys, ys.T)
sigmas[j] /= ws[j,:].sum()
# update complete log likelihoood
ll_new = 0.0
for i in range(n):
s = 0
for j in range(k):
s += pis[j] * mvn(mus[j], sigmas[j]).pdf(xs[i])
ll_new += np.log(s)
if np.abs(ll_new - ll_old) < tol:
break
ll_old = ll_new
return ll_new, pis, mus, sigmas
np.random.seed(123)
# create data set
n = 100
for i in range(0,4):
num1 = np.random.uniform(-500,500)
num2 = np.random.uniform(-500,500)
num3 = np.random.uniform(-500,500)
num4 = np.random.uniform(-500,500)
_mus = np.array([[num1,num2], [num3,num4]])
_sigmas = np.array([[[3, 0], [0, 0.5]], [[1,0],[0,2]]])
_pis = np.array([0.6, 0.4])
xs = np.concatenate([np.random.multivariate_normal(mu, sigma, int(pi*n))
for pi, mu, sigma in zip(_pis, _mus, _sigmas)])
# initial guesses for parameters
num_clusters = 4
pis = np.random.random(num_clusters)
pis /= pis.sum()
mus = np.random.random((num_clusters,2))
print(mus)
sigmas = np.array([np.eye(2)] * num_clusters)
ll1, pis1, mus1, sigmas1 = em_gmm_orig(xs, pis, mus, sigmas)
#Ploy
intervals = 101
ys = np.linspace(-8,8,intervals)
X, Y = np.meshgrid(ys, ys)
_ys = np.vstack([X.ravel(), Y.ravel()]).T
z = np.zeros(len(_ys))
for pi, mu, sigma in zip(pis1, mus1, sigmas1):
z += pi*mvn(mu, sigma).pdf(_ys)
z = z.reshape((intervals, intervals))
ax = plt.subplot(111)
plt.scatter(xs[:,0], xs[:,1], alpha=0.2)
plt.contour(X, Y, z, N=10)
plt.axis([-8,6,-6,8])
ax.axes.set_aspect('equal')
plt.tight_layout()
从scipy.stats导入多变量正常值作为mvn
将numpy作为np导入
将matplotlib.pyplot作为plt导入
def em_gmm_orig(xs,pis,mus,sigmas,tol=0.01,max_iter=100):
n、 p=xs.shape
k=len(pis)
ll_old=0
对于范围内的i(最大值):
exp_A=[]
exp_B=[]
ll_new=0
#电子步
ws=np.零((k,n))
对于范围内的j(len(mus)):
对于范围(n)中的i:
ws[j,i]=pis[j]*mvn(mus[j],sigmas[j]).pdf(xs[i])
ws/=ws.sum(0)
#M步
pis=np.零(k)
对于范围内的j(len(mus)):
对于范围(n)中的i:
pis[j]+=ws[j,i]
pis/=n
mus=np.零((k,p))
对于范围(k)内的j:
对于范围(n)中的i:
mus[j]+=ws[j,i]*xs[i]
mus[j]/=ws[j,:].sum()
sigmas=np.zero((k,p,p))
对于范围(k)内的j:
对于范围(n)中的i:
ys=np.重塑(xs[i]-mus[j],(2,1))
sigmas[j]+=ws[j,i]*np.dot(ys,ys.T)
sigmas[j]/=ws[j,:].sum()
#像日志一样更新完整日志
ll_new=0.0
对于范围(n)中的i:
s=0
对于范围(k)内的j:
s+=pis[j]*mvn(mus[j],sigmas[j]).pdf(xs[i])
ll_new+=np.log
如果np.abs(新-旧)
我通常会遇到以下错误:
Traceback (most recent call last):
File "C:\Usy", line 82, in <module>
ll1, pis1, mus1, sigmas1 = em_gmm_orig(xs, pis, mus, sigmas)
File "C:\Usy", line 48, in em_gmm_orig
s += pis[j] * mvn(mus[j], sigmas[j]).pdf(xs[i])
File "y", line 354, in __call__
allow_singular=allow_singular)
...
ValueError: array must not contain infs or NaNs
回溯(最近一次呼叫最后一次):
文件“C:\Usy”,第82行,在
ll1,pis1,mus1,sigmas1=em_gmm_orig(xs,pis,mus,sigmas)
文件“C:\Usy”,第48行,em\u gmm\u orig
s+=pis[j]*mvn(mus[j],sigmas[j]).pdf(xs[i])
文件“y”,第354行,在调用中__
allow_singular=allow_singular)
...
ValueError:数组不能包含INF或NAN
有什么办法可以解决这个问题吗?最有可能的是,你的sigma值在mvn(mus[j],sigmas[j]).pdf(xs[I])中为零,这给出了奇点。你可能想检查一下你的西格玛值