Keras is a high-level neural network API written in Python. This deep learning Python library can run on top of other source platforms like TensorFlow.
Keras is based on the idea of minimizing the delay between implementing a specific idea and getting a result. The minimization of this delay is the key to good research.
Let’s have a look at what Keras has to offer:
Keras is a user-friendly API that was built with user experience in mind. The API reduces the cognitive load by providing consistent, easily understandable methods as well as clear and meaningful feedback.
Keras is a modular API which means that various schemes like neural layers, cost functions, activation functions, etc. are perfectly standalone modules. This makes the API easier to update and helps to create new models by combining existing modules.
Keras API is easily extensible as existing modules provide sufficient examples for the creation of new modules that are to be used in advanced research.
Perhaps the biggest advantage associated with the Keras API is that the models are described in Python. This makes the API much more robust, easily debuggable, and vastly extensible.
In the example below, we are using Keras
backend module for some common matrix operations.
from keras import backend as bk mat_1 = bk.random_uniform_variable(shape=(2, 2), low = 0, high = 1) mat_2 = bk.random_uniform_variable(shape=(2, 2), low = 5, high = 10) # dot product -- transpose print(bk.dot(mat_1, bk.transpose(mat_2)))
View all Courses