Merge pull request #1 from anubhav2901/patch-1

Patch 1

.github/workflows/main.yml

@@ -0,0 +1,18 @@
+name: Run demo
+
+on: [push]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v3
+      - uses: actions/setup-python@v3
+
+      - name: Install dependencies
+        run: pip install -r requirements.txt
+
+      - name: Run demo.py 
+        run: 
+          python demo.py

demo.py

@@ -0,0 +1,86 @@
+import numpy as np
+import tensorflow as tf
+from keras.layers import Dense
+
+import matplotlib.pyplot as plt
+
+def flood_categorical_crossentropy(y_true, y_pred):
+    loss = tf.keras.losses.categorical_crossentropy(y_true, y_pred)
+    loss = tf.math.abs(loss - b) + b
+    return loss
+
+#testing flooding for MNIST
+
+#import the dataset
+
+print("Testing benefits of Flooding for MLP trained on MNIST dataset\n")
+
+(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()
+
+print("Shape of x_train: {} y_train: {}".format(x_train.shape, y_train.shape))
+print("Shape of x_test: {} y_test: {}".format(x_test.shape, y_test.shape))
+
+
+#data preprocessing
+x_train = x_train.reshape(60000, 784) #reshape 28 x 28 image to 784-length vectors.
+x_test = x_test.reshape(10000, 784)   #reshape 28 x 28 image to 784-length vectors.
+
+x_train = x_train.astype("float32")   #change int to float
+x_test = x_test.astype("float32")
+
+x_train /= 255                        #normalizing
+x_test /= 255
+
+
+num_classes = 1000                           #number of classes
+y_train = tf.one_hot(y_train, num_classes)   #one hot encoding
+y_test = tf.one_hot(y_test, num_classes)
+
+print("Training matrix shape", x_train.shape)
+print("Testing matrix shape", x_test.shape)
+
+
+#define the MLP model
+num_nodes = 1000                      #number of nodes in hidden layers
+batch_size = 200
+
+model = tf.keras.Sequential()
+model.add(Dense(num_nodes, input_shape=(784,), activation="relu"))
+model.add(Dense(num_nodes, activation="relu"))
+model.add(Dense(num_classes, activation="softmax"))
+
+
+#set flood value
+b = 0.01
+
+SGD = tf.keras.optimizers.SGD(learning_rate=0.1, momentum=0.9,)
+model.compile(loss=flood_categorical_crossentropy, optimizer=SGD, metrics=["mse", "acc"])
+history = model.fit(x_train, y_train, batch_size=batch_size, epochs=10, validation_data=(x_test, y_test), verbose=0)
+model.evaluate(x_test,  y_test, verbose=2)
+
+#plot figure
+fig = plt.figure()
+plt.plot(history.history['loss'], label="Train loss")
+plt.plot(history.history['val_loss'], label="Test loss")
+plt.legend()
+plt.show()
+
+#define the MLP model1
+model1 = tf.keras.Sequential()
+model1.add(Dense(num_nodes, input_shape=(784,), activation="relu"))
+model1.add(Dense(num_nodes, activation="relu"))
+model1.add(Dense(num_classes, activation="softmax"))
+
+SGD = tf.keras.optimizers.SGD(learning_rate=0.1, momentum=0.9,)
+model1.compile(loss="categorical_crossentropy", optimizer=SGD, metrics=["mse", "acc"])
+history1 = model1.fit(x_train, y_train, batch_size=batch_size, epochs=10, validation_data=(x_test, y_test), verbose=0)
+model1.evaluate(x_test,  y_test, verbose=2)
+
+fig = plt.figure()
+plt.plot(history1.history['loss'], label="Train loss")
+plt.plot(history1.history['val_loss'], label="Test loss")
+plt.legend()
+plt.show()
+
+print("Testing Loss: ", history1.history['val_loss'][-1])
+print("Testing Loss w/ Flooding: ", history.history['val_loss'][-1])

figures/.gitkeep

@@ -0,0 +1 @@
+

models/mlp_fmnist.py

@@ -0,0 +1,106 @@
+import numpy as np
+import tensorflow as tf
+from keras.layers import Dense
+
+import matplotlib.pyplot as plt
+
+#testing flooding for Fashion-MNIST
+
+#import the dataset
+
+print("Testing benefits of Flooding for MLP trained on Fashion-MNIST dataset\n")
+
+(x_train, y_train), (x_test, y_test) = tf.keras.datasets.fashion_mnist.load_data()
+
+print("Shape of x_train: {} y_train: {}".format(x_train.shape, y_train.shape))
+print("Shape of x_test: {} y_test: {}".format(x_test.shape, y_test.shape))
+
+
+#data preprocessing
+x_train = x_train.reshape(60000, 784) #reshape 28 x 28 image to 784-length vectors.
+x_test = x_test.reshape(10000, 784)   #reshape 28 x 28 image to 784-length vectors.
+
+x_train = x_train.astype("float32")   #change int to float
+x_test = x_test.astype("float32")
+
+x_train /= 255                        #normalizing
+x_test /= 255
+
+
+num_classes = 1000                           #number of classes
+y_train = tf.one_hot(y_train, num_classes)   #one hot encoding
+y_test = tf.one_hot(y_test, num_classes)
+
+print("Training matrix shape", x_train.shape)
+print("Testing matrix shape", x_test.shape)
+
+
+#define the MLP model
+num_nodes = 1000                      #number of nodes in hidden layers
+
+model = tf.keras.Sequential()
+model.add(Dense(num_nodes, input_shape=(784,), activation="relu"))
+model.add(Dense(num_nodes, activation="relu"))
+model.add(Dense(num_classes, activation="softmax"))
+
+
+#set flood value
+b = 0.09                              #selecting value of b from {0.01 .. 0.1}
+
+#for categorical crossentropy
+def flood_categorical_crossentropy(y_true, y_pred):
+    loss = tf.keras.losses.categorical_crossentropy(y_true, y_pred)
+    loss = tf.math.abs(loss - b) + b
+    return loss
+
+SGD = tf.keras.optimizers.SGD(learning_rate=0.1, momentum=0.9,)
+model.compile(loss=flood_loss, optimizer=SGD, metrics=["mse", "acc"])
+history = model.fit(x_train, y_train, epochs=100, validation_data=(x_test, y_test))
+model.evaluate(x_test,  y_test, verbose=2)
+
+#plot loss and accuracy
+'''
+fig = plt.figure()
+plt.plot(history.history['loss'], label="Train loss")
+plt.plot(history.history['val_loss'], label="Test loss")
+plt.legend()
+plt.show()
+'''
+
+#define the MLP model1
+model1 = tf.keras.Sequential()
+model1.add(Dense(num_nodes, input_shape=(784,), activation="relu"))
+model1.add(Dense(num_nodes, activation="relu"))
+model1.add(Dense(num_classes, activation="softmax"))
+
+SGD = tf.keras.optimizers.SGD(learning_rate=0.1, momentum=0.9,)
+model1.compile(loss="categorical_crossentropy", optimizer=SGD, metrics=["mse", "acc"])    #using categorical_crossentropy loss
+history1 = model1.fit(x_train, y_train, epochs=100, validation_data=(x_test, y_test))
+model1.evaluate(x_test,  y_test, verbose=2)
+
+#plot loss and accuracy
+'''
+fig = plt.figure()
+plt.plot(history1.history['loss'], label="Train loss")
+plt.plot(history1.history['val_loss'], label="Test loss")
+plt.legend()
+plt.show()
+'''
+
+#comparing accuracies
+fig = plt.figure()
+plt.plot(history1.history['acc'], label="Training acc")
+plt.plot(history.history['acc'], label="Training acc w/ flooding")
+plt.plot(history1.history['val_acc'], label="Testing acc")
+plt.plot(history.history['val_acc'], label='Tesiting acc w/ flooding')
+plt.legend()
+plt.show()
+
+#comparing losses
+fig = plt.figure()
+plt.plot(history1.history['loss'], label="Training loss")
+plt.plot(history.history['loss'], label="Training loss w/ flooding")
+plt.plot(history1.history['val_loss'], label="Testing loss")
+plt.plot(history.history['val_loss'], label='Testing loss w/ flooding')
+plt.legend()
+plt.show()

models/mlp_kmnist.py

@@ -0,0 +1,117 @@
+import numpy as np
+import tensorflow as tf
+from keras.layers import Dense
+
+import tensorflow_datasets as tfds
+
+import matplotlib.pyplot as plt
+
+#testing flooding for KMNIST
+
+#import the dataset
+
+print("Testing benefits of Flooding for MLP trained on KMNIST dataset\n")
+
+dstrain = tfds.as_numpy(tfds.load('kmnist', split='train', batch_size=-1))
+x_train = dstrain['image']
+y_train = dstrain['label']
+
+
+dstest = tfds.as_numpy(tfds.load('kmnist', split='test', batch_size=-1))
+x_test = dstest['image']
+y_test = dstest['label']
+
+(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()
+
+print("Shape of x_train: {} y_train: {}".format(x_train.shape, y_train.shape))
+print("Shape of x_test: {} y_test: {}".format(x_test.shape, y_test.shape))
+
+
+#data preprocessing
+x_train = x_train.reshape(60000, 784) #reshape 28 x 28 image to 784-length vectors.
+x_test = x_test.reshape(10000, 784)   #reshape 28 x 28 image to 784-length vectors.
+
+x_train = x_train.astype("float32")   #change int to float
+x_test = x_test.astype("float32")
+
+x_train /= 255                        #normalizing
+x_test /= 255
+
+
+num_classes = 1000                           #number of classes
+y_train = tf.one_hot(y_train, num_classes)   #one hot encoding
+y_test = tf.one_hot(y_test, num_classes)
+
+print("Training matrix shape", x_train.shape)
+print("Testing matrix shape", x_test.shape)
+
+
+#define the MLP model
+num_nodes = 1000                      #number of nodes in hidden layers
+
+model = tf.keras.Sequential()
+model.add(Dense(num_nodes, input_shape=(784,), activation="relu"))
+model.add(Dense(num_nodes, activation="relu"))
+model.add(Dense(num_classes, activation="softmax"))
+
+
+#set flood value
+b = 0.01                              #selecting value of b from {0.01 .. 0.1}
+
+#for categorical crossentropy
+def flood_categorical_crossentropy(y_true, y_pred):
+    loss = tf.keras.losses.categorical_crossentropy(y_true, y_pred)
+    loss = tf.math.abs(loss - b) + b
+    return loss
+
+SGD = tf.keras.optimizers.SGD(learning_rate=0.1, momentum=0.9,)
+model.compile(loss=flood_categorical_crossentropy, optimizer=SGD, metrics=["mse", "acc"])
+history = model.fit(x_train, y_train, epochs=100, validation_data=(x_test, y_test))
+model.evaluate(x_test,  y_test, verbose=2)
+
+#plot loss
+'''
+fig = plt.figure()
+plt.plot(history.history['loss'], label="Train loss")
+plt.plot(history.history['val_loss'], label="Test loss")
+plt.legend()
+plt.show()
+'''
+
+#define the MLP model1
+model1 = tf.keras.Sequential()
+model1.add(Dense(num_nodes, input_shape=(784,), activation="relu"))
+model1.add(Dense(num_nodes, activation="relu"))
+model1.add(Dense(num_classes, activation="softmax"))
+
+SGD = tf.keras.optimizers.SGD(learning_rate=0.1, momentum=0.9,)
+model1.compile(loss="categorical_crossentropy", optimizer=SGD, metrics=["mse", "acc"])  #using categorical_crossentropy loss
+history1 = model1.fit(x_train, y_train, epochs=100, validation_data=(x_test, y_test))
+model1.evaluate(x_test,  y_test, verbose=2)
+
+#plot loss
+'''
+fig = plt.figure()
+plt.plot(history1.history['loss'], label="Train loss")
+plt.plot(history1.history['val_loss'], label="Test loss")
+plt.legend()
+plt.show()
+'''
+
+#comparing accuracies
+fig = plt.figure()
+plt.plot(history1.history['acc'], label="Training acc")
+plt.plot(history.history['acc'], label="Training acc w/ flooding")
+plt.plot(history1.history['val_acc'], label="Testing acc")
+plt.plot(history.history['val_acc'], label='Tesiting acc w/ flooding')
+plt.legend()
+plt.show()
+
+#comparing losses
+fig = plt.figure()
+plt.plot(history1.history['loss'], label="Training loss")
+plt.plot(history.history['loss'], label="Training loss w/ flooding")
+plt.plot(history1.history['val_loss'], label="Testing loss")
+plt.plot(history.history['val_loss'], label='Testing loss w/ flooding')
+plt.legend()
+plt.show()