Adjust for library updates

Author: rrlyman

n0_network.py

@@ -30,7 +30,7 @@ def fit(self, truthed_data, nEpochs=5000):
                 summary_str = result[0]
 
                 self._writer.add_summary(summary_str, i)
-                train_accuracy = self._accuracy.eval(feed)    
+                train_accuracy = result[1]   
                 if train_accuracy <= (1.0 - 1e-5  ):
                     perfect_count=10;
                 else:
@@ -39,7 +39,7 @@ def fit(self, truthed_data, nEpochs=5000):
                         break;  
 
                 print ("step %d, training accuracy %g"%(i, train_accuracy),flush=True)
-            self._train_step.run(feed_dict=feed)
+            self._sess.run(self._train_step,feed_dict=feed)
 
 
 

n1_2cnv1fc.py

@@ -105,7 +105,7 @@ def max_pool_2x2(x):
         with tf.name_scope("reshape_x_image") as scope:
             self._x_image = tf.reshape(self._ph.image, [-1,self._nCols,self._nRows,1])
 
-        image_summ = tf.image_summary("x_image", self._x_image)
+        image_summ = tf.summary.image("x_image", self._x_image)
 
         """# ==============================================================================
         
@@ -156,7 +156,9 @@ def max_pool_2x2(x):
 
             # append the features, the 2nd on, that go directly to the fully connected layer
             for i in range(2,truthed_features.num_features ):
-                h_pool2_flat = tf.concat(1, [h_pool2_flat, self._ph[i]])  
+                print(i)
+                print(self._ph[i])
+                h_pool2_flat = tf.concat([h_pool2_flat, self._ph[i]],1)  
             h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)
 
         """# ==============================================================================
@@ -184,7 +186,7 @@ def max_pool_2x2(x):
 
             # 1e-8 added to eliminate the crash of training when taking log of 0
             cross_entropy = -tf.reduce_sum(self._ph[0]*tf.log(y_conv+ 1e-8  ))
-            ce_summ = tf.scalar_summary("cross entropy", cross_entropy)
+            ce_summ = tf.summary.scalar("cross entropy", cross_entropy)
 
         with tf.name_scope("train") as scope:
             self._train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
@@ -194,21 +196,22 @@ def max_pool_2x2(x):
             self._prediction = tf.argmax(y_conv,1)
 
             self._accuracy = tf.reduce_mean(tf.cast(self._correct_prediction, dtype))
-            accuracy_summary = tf.scalar_summary("accuracy", self._accuracy)    
+            accuracy_summary = tf.summary.scalar("accuracy", self._accuracy)    
+            weight_summary = tf.summary.histogram("weights", W_fc2)
         """# ==============================================================================
         
         Start TensorFlow Interactive Session
         
         """# ==============================================================================   
 
         self._sess.run(tf.initialize_all_variables())  
-        self._merged = tf.merge_all_summaries()
+        self._merged = tf.summary.merge_all()
         tm = ""
         tp = datetime.datetime.now().timetuple()
         for i in range(4):
             tm += str(tp[i])+'-'
         tm += str(tp[4])    
-        self._writer = tf.train.SummaryWriter("/tmp/ds_logs/"+ tm, self._sess.graph)
+        self._writer = tf.summary.FileWriter("/tmp/ds_logs/"+ tm, self._sess.graph)
 
         def computeSize(s,tens):
             sumC = 1

n1_2cnv2fc.py

@@ -107,7 +107,7 @@ def max_pool_2x2(x):
         with tf.name_scope("reshape_x_image") as scope:
             self._x_image = tf.reshape(self._ph.image, [-1,self._nCols,self._nRows,1])
 
-        image_summ = tf.image_summary("x_image", self._x_image)
+        image_summ = tf.summary.image("x_image", self._x_image)
 
         """# ==============================================================================
         
@@ -219,7 +219,7 @@ def max_pool_2x2(x):
 
             # 1e-8 added to eliminate the crash of training when taking log of 0
             cross_entropy = -tf.reduce_sum(self._ph[0]*tf.log(y_conv+ 1e-8  ))
-            ce_summ = tf.scalar_summary("cross entropy", cross_entropy)
+            ce_summ = tf.summary.scalar("cross entropy", cross_entropy)
 
         with tf.name_scope("train") as scope:
             self._train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
@@ -229,21 +229,21 @@ def max_pool_2x2(x):
             self._prediction = tf.argmax(y_conv,1)
 
             self._accuracy = tf.reduce_mean(tf.cast(self._correct_prediction, dtype))
-            accuracy_summary = tf.scalar_summary("accuracy", self._accuracy)    
+            accuracy_summary = tf.summary.scalar("accuracy", self._accuracy)    
         """# ==============================================================================
         
         Start TensorFlow Interactive Session
         
         """# ==============================================================================      
 
         self._sess.run(tf.initialize_all_variables())  
-        self._merged = tf.merge_all_summaries()
+        self._merged = tf.summary.merge_all()
         tm = ""
         tp = datetime.datetime.now().timetuple()
         for i in range(4):
             tm += str(tp[i])+'-'
         tm += str(tp[4])    
-        self._writer = tf.train.SummaryWriter("/tmp/ds_logs/"+ tm, self._sess.graph)
+        self._writer = tf.summary.FileWriter("/tmp/ds_logs/"+ tm, self._sess.graph)
 
         def computeSize(s,tens):
             sumC = 1

n1_baseTensorNN.py

@@ -141,11 +141,11 @@ def max_pool_2x2(x):
         with tf.name_scope("xent") as scope:
 
             # 1e-8 added to eliminate the crash of training when taking log of 0
-            cross_entropy = -tf.reduce_sum(self._ph[0]*tf.log(y_conv+ 1e-8  ))
-            ce_summ = tf.scalar_summary("cross entropy", cross_entropy)
+            self._cross_entropy = -tf.reduce_sum(self._ph[0]*tf.log(y_conv+ 1e-8  ))
+            ce_summ = tf.scalar_summary("cross entropy", self._cross_entropy)
 
         with tf.name_scope("train") as scope:
-            self._train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
+            self._train_step = tf.train.AdamOptimizer(1e-4).minimize(self._cross_entropy)
 
         with tf.name_scope("test") as scope:        
             self._correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(self._ph[0],1))
@@ -155,10 +155,10 @@ def max_pool_2x2(x):
             accuracy_summary = tf.scalar_summary("accuracy", self._accuracy)    
         """# ==============================================================================
         
-        Start TensorFlow Interactive Session
+        Start TensorFlow Session
         
         """# ==============================================================================          
-        self._sess = tf.InteractiveSession()          
+        self._sess = tf.Session()          
         self._sess.run(tf.initialize_all_variables())  
         self._merged = tf.merge_all_summaries()
         tm = ""
@@ -215,7 +215,7 @@ def fit(self, truthed_data, nEpochs=5000):
                 summary_str = result[0]
                 #acc = result[1]       
                 self._writer.add_summary(summary_str, i)
-                train_accuracy = self._accuracy.eval(feed)    
+                train_accuracy = result[1]  
                 if train_accuracy <= (1.0 - 1e-5  ):
                     perfect_count=10;
                 else:
@@ -224,7 +224,7 @@ def fit(self, truthed_data, nEpochs=5000):
                         break;  
 
                 print ("step %d, training accuracy %g"%(i, train_accuracy),flush=True)
-            self._train_step.run(feed_dict=feed)
+            self._sess.run(self._train_step.run,feed_dict=feed)
 
 
 

n1_image_to_image.py

@@ -107,7 +107,7 @@ def max_pool_2x2(x):
         with tf.name_scope("reshape_x_image") as scope:
             self._x_image = tf.reshape(self._ph.image, [-1,self._nCols,self._nRows,1])
 
-        image_summ = tf.image_summary("x_image", self._x_image)
+        image_summ = tf.summary.image("x_image", self._x_image)
 
         """# ==============================================================================
         
@@ -218,40 +218,40 @@ def max_pool_2x2(x):
         with tf.name_scope("xent") as scope:
 
             # 1e-8 added to eliminate the crash of training when taking log of 0
-            cross_entropy = -tf.reduce_sum(self._ph[0]*tf.log(y_conv+ 1e-8  ))
+            self._cross_entropy = -tf.reduce_sum(self._ph[0]*tf.log(y_conv+ 1e-8  ))
             #cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(
              #   logits, labels, name='xentropy')            
-            ce_summ = tf.scalar_summary("cross entropy", cross_entropy)
+            ce_summ = tf.summary.scalar("cross entropy", self._cross_entropy)
 
         with tf.name_scope("reshape_x_image2") as scope:
             self._x_image2 = tf.reshape(self._ph[0], [-1,int(self._nCols/2),int(self._nRows/2),1])
 
-        image_summ2 = tf.image_summary("x_image2", self._x_image2)            
+        image_summ2 = tf.summary.image("x_image2", self._x_image2)            
 
         with tf.name_scope("train") as scope:
-            self._train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
+            self._train_step = tf.train.AdamOptimizer(1e-4).minimize(self._cross_entropy)
             #self._train_step = tf.train.GradientDescentOptimizer(0.5).minimize(cross_entropy)            
 
         with tf.name_scope("test") as scope:        
             self._correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(self._ph[0],1))
             self._prediction = tf.argmax(y_conv,1)
 
             self._accuracy = tf.reduce_mean(tf.cast(self._correct_prediction, dtype))
-            accuracy_summary = tf.scalar_summary("accuracy", self._accuracy)    
+            accuracy_summary = tf.summary.scalar("accuracy", self._accuracy)    
         """# ==============================================================================
         
-        Start TensorFlow Interactive Session
+        Start TensorFlow Session
         
         """# ==============================================================================      
 
         self._sess.run(tf.initialize_all_variables())  
-        self._merged = tf.merge_all_summaries()
+        self._merged = tf.summary.merge_all()
         tm = ""
         tp = datetime.datetime.now().timetuple()
         for i in range(4):
             tm += str(tp[i])+'-'
         tm += str(tp[4])    
-        self._writer = tf.train.SummaryWriter("/tmp/ds_logs/"+ tm, self._sess.graph)
+        self._writer = tf.summary.FileWriter("/tmp/ds_logs/"+ tm, self._sess.graph)
 
         def computeSize(s,tens):
             sumC = 1
@@ -319,4 +319,36 @@ def test2(self, truthed_data,  title = ''):
                 ocr_utils.montage(output_images,title='TensorFlow Output Images')         
                 ocr_utils.montage(input_images,title='TensorFlow Input Images')        
 
+    def fit_entropy(self, truthed_data, nEpochs=5000):    
+
+        perfect_count=10
+        for i in range(nEpochs):
+        
+            batch = truthed_data.next_batch(100)
+            # assign feature data to each placeholder
+            # the batch list is returned in the same order as the features requested
+            feed = {self._keep_prob: 0.5}
+            for j in range(truthed_data.num_features):
+                feed[self._ph[j]] = batch[j]  
+                
+            if i%100 == 0:
+
+                feed[self._keep_prob] = 1.0
+                result = self._sess.run([self._merged, self._cross_entropy ], feed_dict=feed)    
+                summary_str = result[0]
+ 
+                self._writer.add_summary(summary_str, i)
+                train_entropy = result[1]  
+                if train_entropy >= (2000 ):
+                    perfect_count=10;
+                else:
+                    perfect_count -= 1
+                    if perfect_count==0:
+                        break;   
+                    
+                print ("step %d, training entropy %g"%(i, train_entropy),flush=True)
+            self._sess.run(self._train_step,feed_dict=feed)
+            
+
+    
 

n1_residual3x4.py

@@ -169,7 +169,7 @@ def computeSize(s,tens):
         with tf.name_scope("xent") as scope:
             # 1e-8 added to eliminate the crash of training when taking log of 0
             cross_entropy = -tf.reduce_sum(self._ph[0]*tf.log(y_conv+1e-8))
-            ce_summ = tf.scalar_summary("cross entropy", cross_entropy)
+            ce_summ = tf.summary.scalar("cross entropy", cross_entropy)
 
         with tf.name_scope("train") as scope:
             self._train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
@@ -179,7 +179,7 @@ def computeSize(s,tens):
             self._prediction = tf.argmax(y_conv,1)            
 
             self._accuracy = tf.reduce_mean(tf.cast(self._correct_prediction, dtype))
-            accuracy_summary = tf.scalar_summary("accuracy", self._accuracy)    
+            accuracy_summary = tf.summary.scalar("accuracy", self._accuracy)    
 
         """# ==============================================================================
         
@@ -188,7 +188,7 @@ def computeSize(s,tens):
         """# ==============================================================================          
 
         self._sess.run(tf.initialize_all_variables())  
-        self._merged = tf.merge_all_summaries()
+        self._merged = tf.summary.merge_all()
         tm = ""
         tp = datetime.datetime.now().timetuple()
         for i in range(4):
@@ -200,7 +200,7 @@ def computeSize(s,tens):
         # tensorboard --logdir '/tmp/ds_logs/'
         # See results on localhost:6006 
 
-        self._writer = tf.train.SummaryWriter("/tmp/ds_logs/"+ tm, self._sess.graph)
+        self._writer = tf.summary.FileWriter("/tmp/ds_logs/"+ tm, self._sess.graph)
 
     def computeSize(s,tens):
         sumC = 1

o3_top_secret_python_box.py

@@ -70,7 +70,8 @@
 from sklearn.metrics import accuracy_score
 from sklearn.decomposition import PCA
 from sklearn.metrics import accuracy_score
-from sklearn.lda import LDA
+#from sklearn.model_selection
+from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
 from sklearn.linear_model import LogisticRegression
 from sklearn.cross_validation import train_test_split
 
@@ -233,7 +234,7 @@ def find_min_max(sums):
 
 
 n_components = 2
-lda = LDA(n_components=n_components)
+lda = LinearDiscriminantAnalysis(n_components=n_components)
 
 X_train_lda = lda.fit_transform(X_train, y_train)
 X_test_lda = lda.transform(X_test)

o4_image_to_image.py

@@ -68,7 +68,7 @@
     #output_feature_list = ['font_one_hot','image','italic','aspect_ratio','upper_case']   
 
     # train the digits 0-9 for all fonts
-    input_filters_dict = {'m_label': range(48,58),'italic':0,'strength':.4}
+    input_filters_dict = {'m_label': [43]+list(range(48,58)),'italic':0,'strength':.4}
     #input_filters_dict = {'font':'BANKGOTHIC','m_label': list(range(48,58)),'italic':0,'strength':.7}    
     #input_filters_dict = {}    
     output_feature_list = ['low_pass_image','image']    
@@ -83,7 +83,7 @@
                                 test_size = .2,
                                 engine_type='tensorflow',dtype=dtype)    
     nn = nnetwork.network(ds.train)
-    nn.fit( ds.train,  nEpochs=5000)  
+    nn.fit_entropy( ds.train,  nEpochs=5000)  
     nn.test2(ds.test)
 
 #     train_a_font(input_filters_dict,  output_feature_list, nEpochs = 50000)    

p115_l1_l2_regularization.py

@@ -100,7 +100,7 @@
 
 def weight_graph(regularization = 'l1'):
     weights, params = [], []
-    for c in np.arange(-4, 6):
+    for c in np.arange(0, 6):
         lr = LogisticRegression(penalty=regularization, C=10**c, random_state=0)
         lr.fit(X_train_std, y_train)
         weights.append(lr.coef_[1])

p119_squential_backward_selection.py

@@ -80,6 +80,7 @@ def __init__(self, estimator, k_features,
         self.k_features = k_features
         self.test_size = test_size
         self.random_state = random_state
+        
     def fit(self, X, y):
         X_train, X_test, y_train, y_test = \
         train_test_split(X, y, test_size=self.test_size,

p141_linear_descriminant_analsys.py

@@ -57,8 +57,8 @@
 import numpy as np
 import ocr_utils
 import matplotlib.pyplot as plt
-from sklearn.lda import LDA
-        
+#from sklearn.lda import LDA
+from sklearn.discriminant_analysis import  LinearDiscriminantAnalysis as LDA   
 print_limit = 20
 chars_to_train = range(48,58)
 columnsXY=range(0,20)
@@ -115,7 +115,7 @@
 print('Within-class scatter matrix: {}x{}'.format(S_W.shape[0], S_W.shape[1]))
 
 print('Class label distribution: %s'
-    % np.bincount(np.array(y_train,dtype='int32'))[min(y_train):])
+    % np.bincount(np.array(y_train,dtype='int32'))[int(min(y_train)):])
 
 d = S_W.shape[1] # number of features
 S_W = np.zeros((d, d))

p177_k_fold_cross_validation.py

@@ -49,7 +49,7 @@
 import matplotlib.pyplot as plt
 import numpy as np
 from sklearn.cross_validation import StratifiedKFold
-from sklearn.lda import LDA
+from sklearn.discriminant_analysis import  LinearDiscriminantAnalysis as LDA  
 
 if __name__ == '__main__':
     #charsToTrain=range(48,58)

p356_neural_net.py

@@ -178,7 +178,7 @@ def _encode_labels(self, y, k):
         """
         onehot = np.zeros((k, y.shape[0]))
         for idx, val in enumerate(y):
-            onehot[val, idx] = 1.0
+            onehot[int(val), idx] = 1.0
         return onehot
 
     def _initialize_weights(self):
@@ -587,7 +587,7 @@ def _encode_labels(self, y, k):
         """
         onehot = np.zeros((k, y.shape[0]))
         for idx, val in enumerate(y):
-            onehot[val, idx] = 1.0
+            onehot[int(val), idx] = 1.0
         return onehot
 
     def _initialize_weights(self):