Keras is an open-source Python library, which by wrapping the numerical computation library TensorFlow, gives its users easy access to deep-learning model programming. It is hosted on Github, compatible with Python 2.7 and 3.6, and its initial release was in 2015.
It was built as part of the research project ONEIROS (Open-ended Neuro-Electronic Intelligent Robot Operating System) with a focus on user-friendliness, modularity, and extensibility. Its main contributor was Google engineer François Chollet.
Keras is part of the TensorFlow 2.0 ecosystem and covers all steps of the machine learning process, from data management to hyper-parameter training to model deployment. It can be used in small research projects or scaled up to an industrial level.
It supports standard, convolutional, and recurrent neural networks; and has other utilities such as dropout, batch normalization, and pooling.
This library is very well documented and has a vital support community.
Keras has gained an essential role in the market. It is used by many scientific organizations around the world, such as CERN, NASA, and NIH because its flexibility allows for initial and high-level experimentation. It is also widely chosen for many university courses on deep learning and by top-performers at Kaggle competitions.
Keras is a powerful library that helps data analysts to quickly develop models. However, with the increasing availability of new modeling algorithms and methods, it is difficult to determine which model is optimal.
LogicPlum’s platform solves this problem through automation, which allows users to test hundreds of different modeling alternatives in a short time, and then choose the best option based on a specified metric.
As a result, creating commercial and research models becomes a task where all participants can contribute, regarding their level of knowledge in mathematics and statistics.
For those wanting to explore Keras:
© 2020 LogicPlum, Inc. All rights reserved.