python计算矩阵的散度_计算类内类间散度矩阵（附代码）

class compute_scatter_matrix(object):

def __init__(self,X,labels):

self.X = X

self.labels = labels

def compute_scatter(self):

meanAll = np.mean(self.X,axis=0)

St = np.dot((self.X-meanAll).T,(self.X-meanAll))

xclasses = {}

meanclasses = {}

Sw = np.zeros((self.X.shape[1],self.X.shape[1]))

for label in np.unique(self.labels):

xclasses[label] = [self.X[i] for i in range(len(self.labels)) if self.labels[i]==label]

meanclasses[label] = np.mean(xclasses[label],axis=0)

Sw += np.dot((xclasses[label]-meanclasses[label]).T,(xclasses[label]-meanclasses[label]))

SB = np.zeros((self.X.shape[1],self.X.shape[1]))

#第二种SB的计算方法

for label in np.unique(self.labels):

n = np.array(xclasses[label]).shape[0]

mean_vec = np.array(meanclasses[label]