Skip to main content

Keras

Keras is an open source deep learning framework for python that runs on top of the machine learning platform Tensorflow. It was developed by Francois Chollet, a Google engineer. It doesn't handle low-level computation but instead, it uses "Backend" library to do it. It is designed to be fast, easy to use and modular. Keras makes Deep Learning so simple. It has multiple tools and functions which allow to work with images, text datasets simpler. Keras not only supports standard neutral network but convolutional and Recurrent neutral network too.

Examples of commonly used models include neural networks to identify handwritten digits using the library, translation of languages from korean to english, precribing and writing antibiotics for infections.

Install Keras :

Install tensorflow using PIP command

$Pip Install Tensorflow
Import Keras

Keras API :

It is placed in hierarchical order with low-level the high-level APIs. The higher end of the API will use a high-level APi named Tf.Keras to train APIs above machine learning models and to make predictions.
Tf.Keras is the tensorflow variant of the open-source library. It was initially released on 27 March 2015 and developed by Various. It was code using python.

Advantages :

  1. It is user friendly, fast in deployment and easy to understand.

  2. It provides large community support as there are lot of AI communities that use Keras for their Deep Learning framework and many publish tutorials and codes to the general public.

  3. It has multiple backends so you can choose a different backend like Tensorflow, CNTK, and Theano to different projects according to your needs as each backend has its own unique advantage.

  4. It is Cross-Platform and easily can be deployed on any device like Cloud engine, Web browser with .js support, iOS with CoreML, Android with Tensorflow Android with a variety of supported devices and platforms.

  5. It support multipe GPUs at once because it has a built-in support for data parallelism.

    Disadvantages :

  6. It cannot handle low-level API so it is not very useful if you are willing to make your own abstract layer for your research purposes because it already has pre-configured layers.

  7. It cause difficulty in debugging error blogs.

  8. It's data pre-processing tools can improve.

  9. Keras has low speed.

    List of modules available in the Keras :

  • Activations - It provides a list of activator function.

  • Text processing - It provides functions to convert text into NumPy array.

  • Image processing - It provides functions to convert images into Numpy array.

  • Sequence processing - Provides functions to generate time based data from the given input data.

  • Bankend - Provides function of the backend library like Theano.

  • Regularizers - Provides a list of regularizers function.

  • Constraints - Provides a list of constraints function.

  • Utilities - It provides a lot of utility function that is useful in deep learning.

    References :

  1. pianalytix
  2. tutorialspoint