Keras, an open source deep learning API

Hard

Keras logo

With the great popularity that the use of artificial intelligence has gained in different areas, “Deep Learning” (deep learning), has also managed to gain great relevance, since it is used for decision making, object detection, speech recognition, language translation and for many more tasks, because it only mentioned some of the ones used.

Black to mention an example, deep learning It is used in surveillance cameras today and in this case we are talking about commercial use, which represents a large market and above all because video surveillance is no longer a luxury, but has begun to become a necessity.

In this way, there is a wide variety of both commercial and open source projects, both for this area of ​​video surveillance and for other use cases of deep learning.

About Keras

Leaving aside the commercial issue and focusing a little more on the title of the article, I would like to talk a little about Keras, which is a High-level neural network API written in Python. This neural network library open source is designed to provide rapid experimentation with deep neural networks and can run on top of CNTK, TensorFlow, and Theano.

As highlights to Keras from other similar projects, is that It is designed so that people can learn it easily, since it was created to be simple, with consistent and simple APIs, it reduces the actions necessary to implement common code and clearly explains user errors.

Hard provides a Python interface with a high level of abstraction and, at the same time, you have the option of multiple servers for calculation purposes. This makes Keras slower than other deep learning frameworks, but extremely beginner-friendly. as it focuses on being modular, easy to use and extensible. It does not handle low-level calculations; instead, it transfers them to another library called Backend.

Another point in favor of Keras is that allows users to produce deep models both iOS, Android, as well as on the web and in JVM, plus it has strong multi-GPU support and distributed training support.

Keras 3

It is worth mentioning that Keras, It is currently on its 3.x branch, which was released a few weeks ago and this new branch has already been receiving some improvements and corrections, with which we can realize that the project is in constant development and that it has a large active community.

Keras has been in intense public beta testing for several months, and the release of Keras 3 represents a complete rewrite, improving the capabilities for training and deploying models at scale.

Within main features of this new branch of Keras 3, the following stands out:

Multi-backend support

Without a doubt one of the great novelties of Keras 3.0 is its unprecedented support for multiple backends, since it acts as a super connector with the ability to dynamically select the backend that will provide the best performance without having to change anything in the code.

Performance improvements

Another of the key highlights of Keras 3.0 is the performance improvements, because it leverages XLA (Accelerated Linear Algebra) compilation to optimize mathematical calculations, in addition to doubling down on performance optimization, integrating techniques such as mixed precision training and distributed training

Expanded ecosystem

With this new update, Keras received support improvements and can be instantiated as PyTorch, exported as a TensorFlow model, or instantiated as a stateless JAX function. This means that you can leverage the strengths of each extended Keras ecosystem framework without being locked into a single ecosystem.

It is worth mentioning that Keras 3 is highly compatible with Keras 2, since it implements the Keras 2 API, with a limited number of exceptions, so most users will not have to make any code changes to start running their scripts. Keras in this new version.

Finally, If you are interested in knowing more about it, you can check the details of this new branch In the following link. If you want to know the how to implement Keras? on your system, you can check the installation methods in this link, while for just like that documentation and use cases To learn about it, you can do it at this link.


Leave a Comment

Your email address will not be published. Required fields are marked with *

*

*

  1. Responsible for the data: AB Internet Networks 2008 SL
  2. Purpose of the data: Control SPAM, comment management.
  3. Legitimation: Your consent
  4. Communication of the data: The data will not be communicated to third parties except by legal obligation.
  5. Data storage: Database hosted by Occentus Networks (EU)
  6. Rights: At any time you can limit, recover and delete your information.