Update myMetrics.py

myMetrics.py

@@ -2,83 +2,6 @@
 import numpy as np
 
 
-def ac(y_true,y_pred):
-    sign = np.zeros((np.max(y_true)+1,np.max(y_true)+1))
-    for i in range(sign.shape[0]):
-        index = np.where(y_pred==i)
-        for j in range(sign.shape[1]):
-            sign[i,j] = np.sum(y_true[index]==j)
-    return sign
-
-def pre(x):
-    n = x.shape[0]
-    min_r = np.min(x,axis = 1)
-    min_r.shape = n,1
-    x -= np.repeat(min_r,n,axis = 1)
-
-    min_l = np.min(x,axis = 0)
-    min_l.shape = 1,n
-    x -= np.repeat(min_l,n,axis = 0)
-    return x
-
-def select_delete(x):
-    sign = 1*(x==0)
-    select = np.zeros(x.shape)
-    delete = np.zeros(x.shape)
-    while np.sum(sign)!=0:
-        nb_zeros = np.sum(x==0,axis = 1)
-        nb_zeros[np.where(nb_zeros==0)]=100000
-        locx = np.argmin(nb_zeros)
-        index = np.where(x[locx,:]==0)
-        locy = index[0][0]
-        select[locx,locy]=1
-        delete[locx,np.where(x[locx,:]==0)]=1
-        delete[np.where(x[:,locy]==0),locy]=1
-        delete[locx,locy]=0
-        x[np.where(delete==1)] = -1
-        x[np.where(select==1)] = -1
-        sign = 1*(x==0)
-    return select,delete
-
-def row_line(x,select,delete):
-    row = np.zeros(x.shape[0],)
-    line = np.zeros(x.shape[0],)
-    row[np.where(np.sum(select,axis = 0)==0)]=1
-    sign_yt = np.unique(np.where(x[np.where(row==1),:]==0))
-    sign_y = sign_yt[0]
-    line_temp = np.copy(line)
-    line_temp[sign_y]=1
-    sign_xt = np.where(x[:,np.where(np.sign(line+line_temp)==1)]==0)
-    sign_x = sign_xt[0]
-    row_temp = np.copy(row)
-    row_temp[sign_x]=1
-    while np.sum(line==np.sign(line+line_temp))+np.sum(row==np.sign(row+np.sign(row+row_temp)))!=2*x.shape[0]:
-        row = np.sign(row+np.sign(row+row_temp))
-        line = np.sign(line+line_temp)
-        row[np.where(np.sum(select,axis = 0)==0)]=1
-        sign_yt = np.unique(np.where(x[np.where(row==1),:]==0))
-        sign_y = sign_yt[0]
-        line_temp = np.copy(line)
-        line_temp[sign_y]=1
-        sign_xt = np.where(x[:,np.where(np.sign(line+line_temp)==1)]==0)
-        sign_x = sign_xt[0]
-        row_temp = np.copy(row)
-        row_temp[sign_x]=1
-    return row,line
-
-def adjust(x,row,line):
-    res = np.copy(x)
-    adjt = x[row==1,:]
-    adjt = adjt[:,line==0]
-    adj = np.min(adjt)
-    for i in range(x.shape[0]):
-        for j in range(x.shape[1]):
-            if (row[i]==1) & (line[j]==0):
-                res[i,j] -= adj
-            if (row[i]==0) & (line[j]==1):
-                res[i,j] += adj
-    return res
-
 def NMI(y_true,y_pred):
     return metrics.normalized_mutual_info_score(y_true, y_pred)
 
@@ -98,26 +21,6 @@ def ACC(y_true,y_pred):
     return sum([w[i,j] for i,j in ind])*1.0/Y_pred.size,ind
 
 if __name__=='__main__':
-    matfn="mnistResult11.mat"
-    f = sio.loadmat(matfn)
-    acc = f['acc']
-    output = f['highOutput']
-    y_true = f['y_true']
-    y_pred = np.argmax(output[10],axis = 1)
-    y_true.shape = 70000
-    print(NMI(y_true,y_pred))
-    print(ARI(y_true,y_pred))
-    print(ACC(y_true,y_pred))
 
-    matfn="stlResult1.mat"
-    f = sio.loadmat(matfn)
-    acc = f['acc']
-    output = f['highOutput']
-    y_true = f['y_true']
-    y_pred = np.argmax(output[2],axis = 1)
-    y_true.shape = 13000
-    print(NMI(y_true,y_pred))
-    print(ARI(y_true,y_pred))
-    print(ACC(y_true,y_pred))