Merge branch 'master' of github.com:lazyprogrammer/machine_learning_examples

Author: Lazy Programmer

ann_class/tf_example.py

@@ -58,7 +58,7 @@ def forward(X, W1, b1, W2, b2):
 logits = forward(tfX, W1, b1, W2, b2)
 
 cost = tf.reduce_mean(
-  tf.nn.softmax_cross_entropy_with_logits(
+  tf.nn.softmax_cross_entropy_with_logits_v2(
     labels=tfY,
     logits=logits
   )

ann_class2/adam.py

@@ -19,13 +19,9 @@ def main():
     max_iter = 10
     print_period = 10
 
-    X, Y = get_normalized_data()
+    Xtrain, Xtest, Ytrain, Ytest = get_normalized_data()
     reg = 0.01
 
-    Xtrain = X[:-1000,]
-    Ytrain = Y[:-1000]
-    Xtest  = X[-1000:,]
-    Ytest  = Y[-1000:]
     Ytrain_ind = y2indicator(Ytrain)
     Ytest_ind = y2indicator(Ytest)
 

ann_class2/batch_norm_tf.py

@@ -183,11 +183,7 @@ def predict(self, X):
 
 def main():
   # step 1: get the data and define all the usual variables
-  X, Y = get_normalized_data()
-  # Xtrain, Xtest, Ytrain, Ytest = train_test_split(X, Y, test_size=0.03)
-  X, Y = shuffle(X, Y)
-  Xtrain, Ytrain = X[:-1000], Y[:-1000]
-  Xtest, Ytest = X[-1000:], Y[-1000:]
+  Xtrain, Xtest, Ytrain, Ytest = get_normalized_data()
 
   ann = ANN([500, 300])
 

ann_class2/batch_norm_theano.py

@@ -202,11 +202,7 @@ def score(self, X, Y):
 
 def main():
   # step 1: get the data and define all the usual variables
-  X, Y = get_normalized_data()
-  # Xtrain, Xtest, Ytrain, Ytest = train_test_split(X, Y, test_size=0.3)
-  X, Y = shuffle(X, Y)
-  Xtrain, Ytrain = X[:-1000], Y[:-1000]
-  Xtest, Ytest = X[-1000:], Y[-1000:]
+  Xtrain, Xtest, Ytrain, Ytest = get_normalized_data()
 
   ann = ANN([500, 300])
   ann.fit(Xtrain, Ytrain, Xtest, Ytest, show_fig=True)

ann_class2/cntk_example.py

@@ -37,22 +37,21 @@
 
 
 # get the data, same as Theano + Tensorflow examples
-X, Y = get_normalized_data()
+Xtrain, Xtest, Ytrain, Ytest = get_normalized_data()
 
 # get shapes
-N, D = X.shape
-K = len(set(Y))
+N, D = Xtrain.shape
+K = len(set(Ytrain))
 
 # we want one-hot encoded labels
-Y = y2indicator(Y)
+Ytrain = y2indicator(Ytrain)
+Ytest = y2indicator(Ytest)
 
 # split the data
 X = X.astype(np.float32)
 Y = Y.astype(np.float32)
-Xtrain = X[:-1000,]
-Ytrain = Y[:-1000]
-Xtest  = X[-1000:,]
-Ytest  = Y[-1000:]
+Xtest = Xtest.astype(np.float32)
+Ytest = Ytest.astype(np.float32)
 
 
 # the model will be a sequence of layers

ann_class2/dropout_tensorflow.py

@@ -33,16 +33,11 @@ def __init__(self, hidden_layer_sizes, p_keep):
         self.hidden_layer_sizes = hidden_layer_sizes
         self.dropout_rates = p_keep
 
-    def fit(self, X, Y, lr=1e-4, mu=0.9, decay=0.9, epochs=15, batch_sz=100, split=True, print_every=20):
-        # make a validation set
-        X, Y = shuffle(X, Y)
+    def fit(self, X, Y, Xvalid, Yvalid, lr=1e-4, mu=0.9, decay=0.9, epochs=15, batch_sz=100, print_every=50):
         X = X.astype(np.float32)
         Y = Y.astype(np.int64)
-        if split:
-            Xvalid, Yvalid = X[-1000:], Y[-1000:]
-            X, Y = X[:-1000], Y[:-1000]
-        else:
-            Xvalid, Yvalid = X, Y
+        Xvalid = Xvalid.astype(np.float32)
+        Yvalid = Yvalid.astype(np.int64)
 
         # initialize hidden layers
         N, D = X.shape
@@ -143,10 +138,10 @@ def relu(a):
 
 def main():
     # step 1: get the data and define all the usual variables
-    X, Y = get_normalized_data()
+    Xtrain, Xtest, Ytrain, Ytest = get_normalized_data()
 
     ann = ANN([500, 300], [0.8, 0.5, 0.5])
-    ann.fit(X, Y)
+    ann.fit(Xtrain, Ytrain, Xtest, Ytest)
 
 
 if __name__ == '__main__':

ann_class2/dropout_theano.py

@@ -39,13 +39,11 @@ def __init__(self, hidden_layer_sizes, p_keep):
         self.hidden_layer_sizes = hidden_layer_sizes
         self.dropout_rates = p_keep
 
-    def fit(self, X, Y, learning_rate=1e-4, mu=0.9, decay=0.9, epochs=8, batch_sz=100, show_fig=False):
-        # make a validation set
-        X, Y = shuffle(X, Y)
+    def fit(self, X, Y, Xvalid, Yvalid, learning_rate=1e-4, mu=0.9, decay=0.9, epochs=8, batch_sz=100, show_fig=False):
         X = X.astype(np.float32)
         Y = Y.astype(np.int32)
-        Xvalid, Yvalid = X[-1000:], Y[-1000:]
-        X, Y = X[:-1000], Y[:-1000]
+        Xvalid = Xvalid.astype(np.float32)
+        Yvalid = Yvalid.astype(np.int32)
 
         self.rng = RandomStreams()
 
@@ -125,7 +123,7 @@ def fit(self, X, Y, learning_rate=1e-4, mu=0.9, decay=0.9, epochs=8, batch_sz=10
 
                 train_op(Xbatch, Ybatch)
 
-                if j % 20 == 0:
+                if j % 50 == 0:
                     c, p = cost_predict_op(Xvalid, Yvalid)
                     costs.append(c)
                     e = error_rate(Yvalid, p)
@@ -166,10 +164,10 @@ def relu(a):
 
 def main():
     # step 1: get the data and define all the usual variables
-    X, Y = get_normalized_data()
+    Xtrain, Xtest, Ytrain, Ytest = get_normalized_data()
 
     ann = ANN([500, 300], [0.8, 0.5, 0.5])
-    ann.fit(X, Y, show_fig=True)
+    ann.fit(Xtrain, Ytrain, Xtest, Ytest, show_fig=True)
 
 
 if __name__ == '__main__':

ann_class2/keras_example.py

@@ -19,17 +19,18 @@
 
 # get the data, same as Theano + Tensorflow examples
 # no need to split now, the fit() function will do it
-X, Y = get_normalized_data()
+Xtrain, Xtest, Ytrain, Ytest = get_normalized_data()
 
 # get shapes
-N, D = X.shape
-K = len(set(Y))
+N, D = Xtrain.shape
+K = len(set(Ytrain))
 
 # by default Keras wants one-hot encoded labels
 # there's another cost function we can use
 # where we can just pass in the integer labels directly
 # just like Tensorflow / Theano
-Y = y2indicator(Y)
+Ytrain = y2indicator(Ytrain)
+Ytest = y2indicator(Ytest)
 
 
 # the model will be a sequence of layers
@@ -60,7 +61,7 @@
 
 
 # gives us back a <keras.callbacks.History object at 0x112e61a90>
-r = model.fit(X, Y, validation_split=0.33, epochs=15, batch_size=32)
+r = model.fit(Xtrain, Ytrain, validation_data=(Xtest, Ytest), epochs=15, batch_size=32)
 print("Returned:", r)
 
 # print the available keys

ann_class2/momentum.py

@@ -26,16 +26,12 @@ def main():
     # 3. batch SGD with Nesterov momentum
 
     max_iter = 20 # make it 30 for sigmoid
-    print_period = 10
+    print_period = 50
 
-    X, Y = get_normalized_data()
+    Xtrain, Xtest, Ytrain, Ytest = get_normalized_data()
     lr = 0.00004
     reg = 0.01
 
-    Xtrain = X[:-1000,]
-    Ytrain = Y[:-1000]
-    Xtest  = X[-1000:,]
-    Ytest  = Y[-1000:]
     Ytrain_ind = y2indicator(Ytrain)
     Ytest_ind = y2indicator(Ytest)
 

ann_class2/mxnet_example.py

@@ -27,18 +27,11 @@
 
 # get the data, same as Theano + Tensorflow examples
 # no need to split now, the fit() function will do it
-X, Y = get_normalized_data()
+Xtrain, Xtest, Ytrain, Ytest = get_normalized_data()
 
 # get shapes
-N, D = X.shape
-K = len(set(Y))
-
-# split the data
-Xtrain = X[:-1000,]
-Ytrain = Y[:-1000]
-Xtest  = X[-1000:,]
-Ytest  = Y[-1000:]
-
+N, D = Xtrain.shape
+K = len(set(Ytrain))
 
 # training config
 batch_size = 32