Question

因此，我正在上一门数据科学课程，这周一直在研究高斯生成分类模型。我编写了以下函数以适合训练数据：

def fit_generative_model(x,y):
    k = 10  # labels 0,1,...,k-1
    d = (x.shape)[1]  # number of features
    mu = np.zeros((k,d))
    sigma = np.zeros((k,d,d))
    sigmac = np.zeros((k,d,d))
    pi = np.zeros(k)
    c = 100
    # identity matrix for sigma
    im = np.identity(d)

    for lb in range(0,k):
        indx = (y == lb)
        mu[lb] = np.mean(x[indx,:], axis=0)
        sigma[lb] = np.cov(x[indx,:], rowvar=0, bias=1)
        #cim = im*c
        #sigmac[lb]=np.matmul(sigma[lb],cim)
        sigmac[lb]=sigma[lb]*c
        pi[lb]=float(sum(indx))/float(len(y))
    # Halt and return parameters
    print(mu.shape,sigma.shape,pi.shape)
    return mu, sigmac, pi

然后我写了：

# let's find out how many errors I have
kp = 10
tp = len(test_labels)

score = np.zeros((tp,kp))
for rw in range(0,tp):
    for pred in range(0,kp):
        score[rw,pred] = np.log(pi[pred]) + \
        multivariate_normal.logpdf(test_data[rw,:], mean = mu[pred,:], cov = sigma[pred,:,:])

predictions = np.argmax(score[:,:], axis=0) + 1
errors = np.sum(predictions != test_labels)
print('Test errors using feature(s):')
for f in features:
    print(featurenames[f],'errors:',str(errors)+'/'+str(tp))

这是我运行它时得到的追溯：

---------------------------------------------------------------------------
LinAlgError                               Traceback (most recent call last)
<ipython-input-15-36dfe1a7193c> in <module>
      7     for pred in range(0,kp):
      8         score[rw,pred] = np.log(pi[pred]) + \
----> 9         multivariate_normal.logpdf(test_data[rw,:], mean = mu[pred,:], cov = sigma[pred,:,:])
     10 
     11 predictions = np.argmax(score[:,:], axis=0) + 1

c:\users\jbustos\appdata\local\programs\python\python37\lib\site-packages\scipy\stats\_multivariate.py in logpdf(self, x, mean, cov, allow_singular)
    493         dim, mean, cov = self._process_parameters(None, mean, cov)
    494         x = self._process_quantiles(x, dim)
--> 495         psd = _PSD(cov, allow_singular=allow_singular)
    496         out = self._logpdf(x, mean, psd.U, psd.log_pdet, psd.rank)
    497         return _squeeze_output(out)

c:\users\jbustos\appdata\local\programs\python\python37\lib\site-packages\scipy\stats\_multivariate.py in __init__(self, M, cond, rcond, lower, check_finite, allow_singular)
    161         d = s[s > eps]
    162         if len(d) < len(s) and not allow_singular:
--> 163             raise np.linalg.LinAlgError('singular matrix')
    164         s_pinv = _pinv_1d(s, eps)
    165         U = np.multiply(u, np.sqrt(s_pinv))

LinAlgError: singular matrix

有人可以帮助我了解我怎么弄错了吗？谢谢。

Python：生成模型LinAlgError：奇异矩阵错误

0 个答案: