Easy and Intuitive Neural Network Library, Keras

Keras is a library that helps in easily creating and training deep learning models and is a popular library when used with TensorFlow.

In simple terms, Keras is a tool created to make TensorFlow easier to use.

Using Keras, you can create a variety of AI models as follows:

• Handwritten Digit Recognition: A model that automatically classifies numbers

• Image Classification: A model that classifies people, cars, cats, etc., in an input image

• Natural Language Processing: A model that assesses whether a news article is positive or negative

• GAN (Generative Adversarial Networks): A model that generates new images

Here is an example of creating a simple AI model.

Copy
from tensorflow import keras
from tensorflow.keras import layers

# Create model
model = keras.Sequential([
    layers.Dense(64, activation='relu', input_shape=(10,)),
    layers.Dense(32, activation='relu'),
    layers.Dense(1, activation='sigmoid')
])

# Compile model
model.compile(optimizer='adam',
              loss='binary_crossentropy',
              metrics=['accuracy'])

# Output model summary
model.summary()

The main points of the above code are as follows:

Don’t worry if you’re unfamiliar with terms like relu, sigmoid, or adam—these will be explained in the Deep Learning Basics chapter.

• from tensorflow import keras: Imports the Keras module from the TensorFlow library.

• from tensorflow.keras import layers: Imports the layers module from the Keras module.

• keras.Sequential(): Builds a deep learning model by stacking neural network layers sequentially.

• Dense(64, activation='relu', input_shape=(10,)): Creates a layer with 64 neurons and uses ReLU as the activation function. The input data is expected to be a one-dimensional vector with 10 features.

• Dense(32, activation='relu'): Creates a layer with 32 neurons and uses ReLU as the activation function.

• Dense(1, activation='sigmoid'): Creates an output layer with 1 neuron and uses Sigmoid as the activation function. Sigmoid converts the output value to a probability between 0 and 1 for binary classification problems.

• model.compile(): Compiles the model. Here, it uses the adam optimizer, binary_crossentropy loss function, and accuracy metric.

• model.summary(): Prints out a summary of the model’s structure.

As shown, Keras enables the easy implementation of complex neural networks using concise and readable code.

In the next lesson, we will learn how to use Keras to train models using real datasets.