Skip to content

Latest commit

 

History

History
115 lines (85 loc) · 3.29 KB

IntroductiontoKeras.md

File metadata and controls

115 lines (85 loc) · 3.29 KB
Raw

Introduction to Keras

Learning Objectives (Competencies)

By the end of this lesson, students will be able to:

  • What are different Deep Learning Platforms in Python
  • Why we use Keras in DS 2.2
  • What is functional and what is sequential API for Keras
  • Apply NN with Keras on iris data

Deep Learning Platforms in Python

1- Keras 2- Tensorflow 3- Pytorch 4- Caffe 5- Theano 6- CNTK 7- MXNET

Why we use Keras in DS 2.2 ?

  • A focus on user experience, easy to build and train a deep learning model
  • Easy to learn and easy to use
  • Large adoption in the industry and research community
  • Multi-backend, multi-platform
  • Easy productization of models

Keras has two API Styles

The Sequential API

  • Dead simple
  • Only for single-input, single-output, sequential layer stacks
  • Good for 70+% of use cases

The functional API

  • Like playing with Lego bricks
  • Multi-input, multi-output, arbitrary static graph topologies
  • Good for 95% of use cases

Activity: Apply NN with Keras on iris data

  • Use Sequential API for Keras
  • Use 70 percent of data for train
  • Use one-hot encoding for labels with from keras.utils import np_utils
  • Define two layers fully connected network
  • Define categorical_crossentropy as the loss (cost) function
from keras.models import Sequential
from keras.layers.core import Dense, Activation
from keras.utils import np_utils
from sklearn.preprocessing import LabelEncoder
from sklearn import datasets
from sklearn.model_selection import train_test_split

iris = datasets.load_iris()
X, y = iris.data, iris.target

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)

y_train_one_hot = np_utils.to_categorical(y_train)
y_test_one_hot = np_utils.to_categorical(y_test)

# print(y_one_hot)

model = Sequential()
model.add(Dense(16, input_shape=(4,)))
model.add(Activation('sigmoid'))
model.add(Dense(3))
model.add(Activation('softmax'))
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=["accuracy"])
model.fit(X_train, y_train_one_hot, epochs=100, batch_size=1, verbose=0);
loss, accuracy = model.evaluate(X_test, y_test_one_hot, verbose=0)
print("Accuracy = {:.2f}".format(accuracy))

Activity: Apply NN with Keras on iris data with Functional API

from keras.layers import Input, Dense
from keras.models import Model
from keras.utils import np_utils
from sklearn.preprocessing import LabelEncoder
from sklearn import datasets
from sklearn.model_selection import train_test_split

iris = datasets.load_iris()
X, y = iris.data, iris.target

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)

y_train_one_hot = np_utils.to_categorical(y_train)
y_test_one_hot = np_utils.to_categorical(y_test)

# print(y_one_hot)

inp = Input(shape=(4,))
x = Dense(16, activation='sigmoid')(inp)
out = Dense(3, activation='softmax')(x)
model = Model(inputs=inp, outputs= out)
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=["accuracy"])
model.fit(X_train, y_train_one_hot, epochs=100, batch_size=1, verbose=0);
loss, accuracy = model.evaluate(X_test, y_test_one_hot, verbose=0)
print("Accuracy = {:.2f}".format(accuracy))